Adaptive Hypermedia

1.1 Educational Adaptive System (EAHS)

Vannevar Bush proposed Hypertext in 1945 [1]. It is considered as non-sequential text, which displayed on the web sites. Pages or nodes are connected with links called hyperlinks. Since the beginning of World Wide Web, Hypertext has become the standard way to communicate information among computers connected to internet. Hypertext has no media within; it is just a text [2]. Using Hypertext users can jump from one page to another by following a link between them. As the world-wide-web developed by time, new media started to be embedded in the web site and some of them are used as hyperlinks, we started to see web sites with audio, video, pictures and animation beside the Hypertext. Hypertext design is shown in figure 1.1.

Figure 1.1: Hypertext design

Later the term Hypertext alone has become not valid anymore because the existence of multimedia. Multimedia means using more than one media in the same presentation to help users in better understanding the presentation concept. The Term Hypermedia replaced the term Hypertext, which composed of Hypertext + multimedia. Hypermedia in the cyber space either static or dynamic, Static Hypermedia gives access to the same page content and same links to all users, each time they visit the site, while the dynamic Hypermedia often change the content and the links [3]. One of the most known dynamic Hypermedia is the adaptive Hypermedia. The researches and review process in last decade of Hypermedia and multimedia provided rich information and technologies for theoretical and practical development of Hypermedia, Hypertext, interactive multimedia and other related technologies. The new systems in this field provided integration between different media, pictures, sound, video and audio and strong retrieval tools [3]. Hypermedia design is shown in figure 1.2.

Adaptive Hypermedia[N2] in its turn is a new version of Hypermedia. As mentioned above it is a dynamic Hypermedia, its functionality is dealing with the topic of adaptation and the topic of Hypermedia. The difference between Hypermedia and adaptive Hypermedia is illustrated in figures 1.2 and 1.3 [4], [5].

The concept of adaptive Hypermedia research was originated in the early 1990s. It takes his importance from the diversity of the users; some users may understand a topic more than other according to the previous background. In the linear book, this is not important since the reader follow the book sections one by one while in Hypermedia it is a problem because we do not know where the reader may go through the web nodes. We therefore introduce the term personalization through the adaptive Hypermedia, which is the process of presenting information in a way suitable for each individual user. A personalized information source should present different information to different users. Normal Hypertext does not really provide personalized information. However, it cannot automatically guide the user to the version that is most suited for him or her. A new generation of Hypermedia systems Adaptive Hypermedia Systems (AHS) has been invented with the purpose of solving these problems.

Figure 1.2: Traditional Hypermedia

The most popular application area for adaptive Hypermedia research is Adaptive Educational Hypermedia Systems (AEHS)[N18]. The goal of the user (learner) is usually to learn all the learning material or a reasonably large part of it through the best suitable way to him.

Figure 1.3: Adaptive Hypermedia

Adaptation aspects should include lectures presentation to a particular student according to his cognitive style, assessment tools that should follow different paths for each student, and finally adaptation should include the one to one chat room discussions among students.

1.2 Teaching and Learning Through Chat

In recent learning systems, technology strongly involved in the pedagogical process, especially in the virtual classroom, virtual classroom contains mainly two components , first is the lectures presentations , second is the student evaluation techniques. Most learning systems manage these two processes in traditional way. Both lectures presentation and evaluation process are introduced in the same way for all students. Personal differences among students specially student preferable way of study are not considered in these systems.

Data about student is needed to build the student model and estimate the student knowledge level and learning style [6]. This evaluation will be reflected to the adaptivity of the learning objects presentation. The framework system AVCM has its technical solutions for adaptive learning based on two sources of student data .The questionnaire answered by the student at the beginning of the course, which will give an initial stereotype for the student, needing to have initial data about the student is an important factor to tell the system how to deal initially with each student in private manner. Later the system should be able to update the student stereotype dynamically during the learning process. The exam follows each concept presentation is used to evaluate the student knowledge and the new cognitive style. An integrated model for adaptive e-learning system is an important factor. It's important to keep models in the system separated from the system processing. Data storage, domain and user models are kept separately from the system to allow other systems which need to use SM and DM to use them independently. Educational process within the virtual classroom can take advantage from the student modeling process by adding new component which is needed for the collaboration and discussion among students [7].The chat room is a strong component used in virtual classroom which enables the student to add value to his knowledge level by discussion with other students. But we still in the same field , the problem of open learning system which doesn't enable the students to know each other's personally and consequently will be difficult to select peer in discussion who can add value to the student knowledge level. AVCM introduced new component to called adaptive chat tool will be discussed in the following chapter. In our research we will discuss this model to build our view in student modeling for AVCM chat.

Teachers who are interested in using chat in their educational activities should know more about the abilities and characteristics of the chat . They could decide better about their activities and tasks during the chat if they understand the role of the chat in education. The experts could be a mail player in designing the educational process to be held in chat room [N5].

1.3 Our Problem and Mission

In AVCM a new model of adaptive chat too is supposed to be integrated with adaptive educational systems. Adaptive chat tool help the students to select their best peers in discussion by marking the students in the chat window with different colors to differentiate among them. Our problem is to model the student within this chat tool in order to reflect the knowledge level acquired with the chat session into the overall student profile. The educational process within the chat tool consists of peer to peer discussion session about one and only one concept. This concept should be first studied within the virtual classroom presentations and also having the adaptive test which give the student a score for this concept. This score is only one input for the student model about this concept. We have other inputs including the student evaluation within the chat session. The evaluation process will be done using three methodologies.

First: Peer evaluation.

Second: Chat context analysis

Third: Time spent in the chat.

Our mission is to build a student model approach for the educational chat tool which can be used to update the student profile and reflect the new entries on the adaptive system.

The study aims to answer the following questions:

Q1: why we need to use adaptive chat in virtual classroom, and what are the technologies used for educational chat.

Q2: What is the integration model between adaptive virtual classroom and chat model?

Q3: How we can evaluate and model the student within the chat tool.

Q4: How we can update the student model using input entries from the chat tool.

1.4 Novelty and Contribution

'AVCM' is a new model in adaptive virtual classrooms. We are talking about new brand of technology and methodology. Since AVCM is a new model then the value added by our research is also new in this field .The table 1.1 shown below describes the comparison results between AVCM chat model and other similar systems and shows the new contribution of our research .

Table 1.1: [8] Position of AVCM among the other similar systems in the world

Our research has strong impact on the educational chat process. In sense it controls the chat session and doesn't leave the chat disoriented among the students because the chat will be evaluated and reflected to the student model. The chat session in this case will not be considered as only assistant to understand the concept but also as a complete educational model.

Chapter2

Student Modeling in Adaptive Educational Chat Room theoretical framework

2.1 Brief Introduction

In this chapter we are going to discuss the background of our research. Our research will be integrated with AVCM project. Our contributions are considered as additional function that helps the AVCM project to enhance the student modeling facility. The educational chat is the main concept which will be the orientation of our research while the evaluation process is necessary to update the student model. In this chapter we are going to survey several related fields. AVCM virtual classroom model, the technologies used in educational chat, and also the dynamic online peer evaluation, the technologies used to calculate the time spent on one session; finally we need a technology to analyze the text in the chat palette to extract the importance of the discussion and the relation with the concept of discussion.

2.2 Theoretical framework

2.2.1 AVCM and Student Modeling

As we discussed in the introduction, AVCM is a model for adaptive e-learning system this model consists of three main facilities:

1. Adaptive presentation

2. Adaptive testing

3. Adaptive chat.

Several models underlined these models to perform the functions required by AVCM as described in [8]:

1) Domain Model (DM)

2) Student Model (SM)

3) Activity Model (AM)

4) Nodes Selector

5) Concept Score Evaluator

6) Cognitive style evaluator

7) Chat room Interface adapter

8) Peer evaluation

A. Domain Model

Domain model is one component of the intelligent tutoring system (ITS)[N19].Domain Model in as described in [9] is a set of chapters (CH), each chapter consists of a set of Modules (M), module is separated into several unites (U) and unit to several subunits (SU), the subunit has several concepts which represent the learning objects of this subunit, the student should follow these concepts one by one in sequence to achieve knowledge required for this subunit .Each concept should be accompanied with expert estimation of the required knowledge to pass this concept in addition to the cognitive style related to this concept .In such a way we can define the concept as the set < name, path, expert level, cgs (ver, vis, lis) >. For example, we can define abacus concept in 101 course as the tuble < abacus, intro counting machines, 77%, ver >.

Expert knowledge serves to decide if the student passes the concept or should be keeping studying the same concept. Cognitive style serves to decide how to present the concept LOs to the student. The Domain Model does not have any teaching materials .It is only describes the structure and the relationships between course components. Domain Model will provide information to several modules in the system; Nodes selector, Question selector, Chat room interface adapter. These modules will be discussed later in this section.

B. Student Model

User modeling is usually referred to the Allen, Cohen and Perrault [N6]. The concept means collection of different types of information about, and exhibited different kinds of adaptation to, their current users

Student Model is the core of the learning adaptive systems and ITS [N17]. It will provide the required data for adaptation techniques beside the Domain Model for both presentation and exam agent .Student Model actually consist of several sources of data, AVCM concentrated on two sources:

' Static Student Modeling (Stereotype):

A stereotype is a popular belief about specific types of individuals. Stereotypes are standardized and simplified conceptions of groups based on some prior assumptions. In AVCM, there are several characteristics and preferences can be considered in the system, which affect the behavior of the system. These preferences are called domain independent. This data mostly static and not changeable during the instructional process, for example and not all AVCM considers preferable time of learning within the day or night, sex, age, background, skills, etc. This data can be gathered directly from the student via a questionnaire, and used to classify the students into groups called stereotypes. Then store this data in the student profile database.

' Dynamic Student Modeling (Overlay Model):

Once the student starts to use the system, his data will be gathered using dynamic modeling. Dynamic modeling has two entries, one related to student the knowledge estimation through exam evaluation and peer evaluation. The other is related to the student new cognitive style. These both modeling processes are described in the following sections.

C. Activity Model (AM)

AM will be the second process right after the student login to the system; for each concept there are two or more activities. AVCM used only two activities, first are the Presentation Agent (PA), second is Exam Agent (EA). PA will have the function of generating the query to retrieve the appropriate nodes to be built the LOs presentation for the current concept and passing this query to the Nodes selector to retrieve the required nodes. Then gives an order to the resource model to adjust the selected nodes and present them as pages to the student.

D. Resource Model

RM consists of forms, which should be built dynamically according to the order of the PA and EA. These forms will be as the result of the queries matched with the parameters passed by PA and EA to QS and NS.

E. Nodes Selector

Nodes selector is a query executer, which receive the query from PA and execute it.

F. Question Selector

Question selector is also receiving a query from EA and executes it.

G. Cognitive style evaluator

Cognitive style evaluator is another accumulative dynamic Student Modeling. It evaluates the new student cognitive style according to his questions answers.

H. Chat room Interface adapter

Chat room interface adapter is an additional separated system, which can get advantage from Student Modeling in AVCM. Adaptive chat room is a collaborative tool for students when they want to discuss among each other. This tool gives an evidence about the student knowledge level for each student participate in chat room relying on student-concept evaluation. In order to give this evidence the chat room takes an input from both student and Domain Model, and adapts the chat room according to the input. The student who has knowledge less than the current student will appear in red color. The student who has knowledge equals to the current student knowledge will be given a yellow color, and finally the student who has higher knowledge will have green color. This way makes it easier for the current student to determine with which student he should discuss the current concept.

I. Peer evaluation

Actually, the student will be evaluated for each concept separately through the exam after each concept learning process. For example, student X got a score of 60% in concept C1 evaluation. During discussion with peers he may have, new knowledge about the concept. This increment in knowledge level should be measured. Once two peers finishing their discussions about one concept, a questionnaire should be popped up for both peers to evaluate each others, after filling the questionnaire it will be analyzed and passed to the Student Model to extract the increment in the students' knowledge and store it in the student profile [8].

The overall architecture of AVCM is shown in figure 2.1

2.2.2 Educational Chat (instant messaging) Technology

Instant Messaging (IM) is a peer-to-peer service for communication with others[N10], which usually includes many chat clients, programs and a central server. Client software enables users to use direct links for communication through the servers. Many instant messaging systems are available these days. The most popular are MSN Messenger [N7], and Yahoo Messenger [N8], and ICQ AOL [N9]. These systems are largely used and they have privacy and comfort [10].

Education chat or instant messaging allows students to have a discussion in real time over the Internet. This is a useful way to share and discuss among peers and classmates. The chat room is much different from forums. Chat has some other features, which make it stronger than forums .The discussion using chat rooms used in open universities in general. Students who feel shy to discuss face to face may find it easier for them to discuss using chat room. So is commonly used in universities that have e-learning systems. The difference between the discussion in face-to-face classroom and virtual classroom is shown in figure 2.2 and 2.3.

Figure 2.2: Traditional discussion with teacher

Vasileios C. in his research about using chat by Greek students made a study about the intensively of using chat room for many purposes [11]. Table 2.1 show that there is a good rate of using the chat for education, but not enough. Many students never use the chat for education and others use it rarely. The reason for this result is the interesting in using chat room for education while students are interested to use it for other purposes. If we want to improve these results, we need to make changes on the chat room to make it more interesting for education. James M. Hudson in [12] made statistical studies about chat and e-learning. One important result he found related to the time length spent in discussion. He found that students spend more time in online conversation than face to face conversation. The results are shown in table 2.3.

.

Figure 2.3: Online discussion with teacher and peers

Table 2.1: Interest in potential uses of chat [11]

Activity Never Rarely Every month Every week Every day

Communicate with friends who live far from you 2 5 14 18 71

Make new friends 24 32 3 10 31

Date 72 20 4 2 2

Educational issues 29 32 18 19 2

Help / Technical support 45 33 17 4 1

Replace phone calls 46 23 9 9 13

Table 2.2: Length of Conversations [12]

Average On-Topic Conversation Length

Face-to-Face 20.68 minutes (StDev = 7.29)

Online 46.27 minutes (StDev = 15.34)

Internet-based messaging based on communication protocols [13]. Instant messaging usually called messenger in common language. Most messengers have many features, which support the communication ability. Some of them are text based only while others support voice and video communication. We may invest the ability of instant messaging systems in our proposed chat room to support the educational process by allowing the students to use it for discussion. Instant messaging systems consist of client side and server side connected through the internet or through the LAN network within the organization. Several users can access the server at the same time. However, in education systems we usually need to limit the numbers of students as much as their number. Figure 2.4 shows the basic architecture of the IM system. Linan Zheng in [13] made a comparison between several systems of IM. This comparison is described in table 2.3.

Figure 2.4: Basic Architecture of IM system

In figure 2.5, we can see the interface of Jabber IM system as an example for instant messaging.

Table 2.3: IM systems comparison [13]

System Advantage Disadvantage

Jabber open-source XML-based decentralized server architecture More or less a concept, not complete and so popular

ICQ send offline message provides great functionalities centralized server architecture not an open-source protocol security aspect synchronous communication leads to slower many operations on client-side

MSN

Messenger asynchronous protocol (SIMPLE) many functionalities centralized server architecture not an open-source protocol Integrates heavily with Windows target for viruses/hackers

Yahoo!

Messenger centralized server architecture excellent integration with Web powerful chat room not an open-source protocol many distractions for user: games, advertising, radio weak security features

Figure 2.5: Jabber messenger

Moodle as framework for e-learning is used widely in open universities [14]. It has a chat utility for direct meeting among students and teachers similar to other instant messaging system like MSN or iChat. In Moodle even if you've set chat times, the chat is always open to students. Moodle does not restrict access to the chat based on the times you set when you create the chat. Instead, it creates entries in the course calendar that remind people to log in for the chat at certain times. Chat in Moodle has mainly two facilities. First, typing messages and broadcast to everyone logged into the chat. Second, beep users. It is possible to create one session for the entire course or set up repeating sessions for multiple meetings. Chat in Moodle is shown in figure 2.6.

Figure 2.6: Moodle chat interface

Moodle chat is widely used in universities [14]. It is a useful technique used for direct meeting in e learning. However, Moodle could be more useful in hybrid learning or blended learning, which mixed between traditional education in classrooms and e learning. Students in such situation may know each other in real so it is acceptable to open a discussion with one who you know before. What we are talking about is open learning, which does not support real meeting in classrooms. In this case the students don't know each other and normally they can't estimate who is the best student to discuss with. Moodle shows all the students logged on the chat room equally same as yahoo and msn. It is possible to use Moodle as framework for our solution but need to be upgraded by our view.

Moodle does not support real adaptive discussion meeting in classrooms. In this case the students don't know each other and normally they can't estimate who is the best student to discuss with. At the same time Moodle has no evaluation process for the student knowledge acquirement in the educational chat room. MOODLE has many weakness points:

1. No time calculation for the chat session

2. No Peer evaluation in Moodle chat room which is very important from our perspective to model the student.

3. No chat text chat analysis in Moodle chats. Therefore we can't extract the rate of the relation between the chat text and the concept of discussion.

In our research study we are going to solve some of these problems by our educational chat tool facilities.

2.2.3 Text Mining and Analysis

Data mining in general is the process to extract patterns from large amount of data .The process relies on AI methods to extract the information and change it to manageable and useful to use. It uses database management systems, inference, visualization and other IT concepts to perform it tasks [N11].

In peer to peer chat process, both parties write number of messages .These messages usually related to one or more concept. In our case we are looking for methodology to analyze the text entered in the chat room to evaluate its relation with the current of discussion .Chat posts usually classified as being a part of conversation or free topic chat , as being about a particular topic [15] , collaborative task 'oriented, academic seminar or presentation chat, practice chat and evaluation chat. We are setting on collaborative ' task oriented chat and evaluation chat .We are interested in the second type because we are looking for estimating the degree of the relation of the chat posts and the current educational concept.

In chat posts we face many chat characteristics which we should take in consideration when we analyze it [10]:

1. Chat text and its component

2. Chat inistialisms which are abbreviations and acronyms used in chat which are used to express some words instead of spoken language.

3. Emoticon usage: Some symbols used to express emotions.

4. Abbreviated speech: Misused of words are more common in chat room.

5. Mentions: Sounds like buzz and poke.

6. Using of different language.

Our objective is to extract the maximum amount of conversation related to the current concept, therefore we find ourselves need to remove many items from the chat context which are not useful to our modeling process. In this case, data mining will be applied to get the minimum and most important data related to the concept, which means that all chat items not useful should be searched and removed from the chat mining. And leave only the text related directly to the concept.

Currently we have number of monitoring systems mainly divided into two categories [16]: network based and client based. In our model network is the best solution since we need the analysis to be centralized and confidential. Table 2.4 shows a comparison of instant messaging monitoring capabilities.

Table 2.4: IM Monitoring Capabilities

Chat analysis in most systems aims to provide the minimum required data by filtering the target text. Most monitoring systems [16] do that by log browsing, message retrieval, content search and simple statistical reporting. In our research we need o use the last three while we are not going to use log browsing. When we are talking about chat in education we mean some specific kind of chat used for educational purpose only.

Hui, He and Dong in [16], proposed chat monitoring system called IM analysis. It uses text mining techniques for chat text analysis .Procedures introduced on chat are for retrieval for general browsing .It discovers the social interactions of IM users with their contacts in addition to topic analysis for the current chat . Our research is interested in this type of analysis of chat text.

Figure 2.7 shows the general methodology used for data mining and analysis of text mining solutions. Metadata showed in the figure 2.7 is a data about the data content which can be implemented using data warehouse and data store. Text mining usually has three steps, first structuring the text entered by the user, second extracting patters from the structured data, and finally evaluation the interpreted output. 'High quality' in text mining usually refers to some combination of relevance, novelty, and interestingness. Typical text mining [16] tasks include:

1. Text categorization.

2. Text clustering.

3. Concept/entity extraction.

4. Production of granular taxonomies,.

5. Sentiment analysis.

6. Document summarization.

7. Entity relation modeling (i.e., learning relations between named entities).

Figure 2.7: Text mining solution [17]

Information retrieval is the key factor of the text mining; this ability can be done using lexical analysis and studying the frequency of words occurrences within the text. We may also make pattern recognition. The most known data mining techniques are:

1. Link and association analysis.

2. Visualization and predictive analytics.

2.2.4 Dynamic Online Peer Evaluation

Peer evaluation process somewhat like online survey .As described in [N13] online survey has good advantages if designed and conducted in proper way .Peer evaluation and online survey are a growing filed over the internet as tools for gathering data from different sources.

'Peer evaluation usually done by colleagues or peers, of all teaching related activities for either formative or summative purposes. It includes course materials, student evaluations, course portfolios, teaching portfolios, documentation of teaching philosophy, teacher self-assessments, classroom observations, and other activities which may be appropriate to a discipline[N12].

Actually, the student will be evaluated for each concept separately through the exam after each concept in learning process. For example, student X got a score of 60% in concept C1 evaluation. During discussion with peers he may have, new knowledge about the concept. This increment in knowledge level should be measured. Once two peers finishing their discussions about one concept, a questionnaire should be popped up as seen in figure 2.8 for both peers to evaluate each others, after filling the questionnaire it will be analyzed and passed to the Student Model to extract the increment in the students' knowledge and store it in the student profile [8].

Figure 2.8: Example of peer evaluation process

The peer evaluation is not more than ordinary questionnaire with multiple choice answers and the students give their own opinion freely. Several techniques and methodology used for online questionnaire or moreover for online survey because our peer evaluation methodology will match the same methodology used for such survey. In [18] online survey is described with multiple choice answers where respondents select one or more options. Online survey which represents the base for our online peer evaluation tends to focus in one more on 'Quantitative' data collection [19] so it is preferable to consider the following points when we design the peer evaluation form.

1. Sate precisely the objectives of the evaluation and what are we going to measure and discover. And consequently the actions required by us to perform the survey.

2. Sate precisely what will the output report look like? What charts and graphs will be prepared? What information do we need to be assured that action is warranted?

3. Give a rank each topic and list the most important topics first.

4. We should anticipate how easy or difficult is it for the respondent to provide information on each topic, and find the easiest way to obtain the information by asking other questions.

Peer assessment is an important process within this research .We will use it as described in the earlier sections in this research. [20] It is an evaluation method that gives an interactive interpretation of the student's knowledge and activities, and to enable the peers to improve their understanding of the others performance. Students develop their skills and enhance thinking through peer assessment. Peer assessment is a form of one to one evaluation, which can include student understanding of the educational environment, and behaves according to this understanding. May be awarding grades by students may not accurately reflect the achievement of the peer because of the autonomy of the individual, and the ability of bias or adequate assessment. But it still one way of student modeling within the chat session. Proposed methodology can be used for extracting the student learning styles who give signs in the evaluation process in order to enhance the accuracy of the assessment. Thus, the assessment of the assembled to consider the learning styles of the students provides students with best feedback.

What are we looking for peer evaluation process are the following points:

1. The student estimation about the peer knowledge level which represents one input of this student score.

2. Peer activities within the chat session modeling, this include the activities of text and time usage.

3. Help the student to know more about his peer by reviewing the last peer assessment of the student.

All these requirements will be discussed in detail in chapter 4.

2.2.5 Time spent on web site calculation

Time is a very important factor in chat session time .Managing the time means more probability to achieve our goals in better way. Time management as described in [N14] 'is a set of principles, practices, skills, tools and systems that help us use time to accomplish what we want'. We need time management in chat session to let the students achieving the maximum benefits from the chat session.

Many web sites have their own way of calculation the time spent on the web site, usually by using login and logout time .How to Interpret Time on Site was discussed by Ross in [21]. He described that time spent provides a better measurement of users' involvement with the site, especially in advertising show. 'The new information will give advertisers, investors and analysts a better understanding of sites' popularity', he said. Time spent on web site will immediately affect some sites, such as Google, which will fall in the rankings because people typically visit often; if the time spent becomes smaller the rank will be minor.

Time spent calculation is an important factor in our research. How long a student stays in educational chat room has long been considered a key indication of how successful that chat room is in attracting relevant students [22]. The theory being that the longer someone spends on the chat room, the more interested they are in what you have to offer and the more he is serious in what he does. Time on chat room can be an indication of the level of interest or involvement that a student has with the chat room. It is also a good indicator of the success of what the student is doing on the chat room [23].

Time spent on chat room has two important aspects:

1- Session time: Login and log out .The total time spent on the chat room.

2- Activity time: The average time of activities done by the student.

Almost universally when we work with a client on the issue of time on chat room the question is raised; well what if I leave my chat room open over the session time , does that impact how long the web analytics tool shows as having spent time on the site and consequently the evaluation process? These questions are needed to be considered in our research because we are going to consider many aspects of time activities within the chat room. For example how is the frequency of posting messages by the student's .How long time the student leaves the chat without any activity? And many other aspects related to time modeling in the educational chat room.

2.2.6 Regression analysis in statistics

The role of using statistics specially in science and engineering is an important factor in finding results. [N15] described the role of statistics in experiments is described in figure 2.8a.

3

Figure 2.8a Critical stages of statistical input in scientific investigations

The main output shown in the last figure is the observation, which we need exactly from our experiments. Regression analysis is one tool of statistical analysis which we will use it mainly in our research.

Regression analysis is the process to estimate the relationship among variables [24] .It has many different techniques for modeling and analysis the variables, when we concentrate on the relation between the dependent and or o more independent variables. Regression analysis helps one understand how the typical value of the dependent variable (or 'Criterion Variable') changes when any one of the independent variables is varied, while the other independent variables are held fixed. Regression analysis finds the l expectation of the dependent variable if it knows the independent variables. In all cases, the estimation target is a function of the independent variables called the regression function. In regression analysis, it is also of interest to characterize the variation of the dependent variable around the regression function which can be described by a probability distribution.

Regression analysis is a useful tool for predicting and discovering new parameters which can be used in machine learning and computer applications [25]. Sometimes regression analysis is used to find the best independent variable related to the dependent variable .Regression analysis can be used to infer causal relationships between the independent and dependent variables. However this can lead to illusions or false relationships, so caution is advisable [25]. Figure 2.9 shows the regression analysis example [26].

Figure 2.9: Regression analysis example

Linear regression and ordinary least squares regression are parametric use a finite number of unknown parameters that are extracted from the data. Nonparametric regression refers to techniques that allow the regression function to give parameters in its own way. Regression models usually works with moderated assumptions, so it is generally useful and can make optimal results. But in some cases especially with small effects or questions of causality , regression methods can give misleading results [27].

2.2.7 Rule-based system facilities

Using AI in educational systems is required to develop an AIED system able to detect and then appropriately react to an affective state of a learner[N16].Because we are talking about intelligent learning environment we have several approaches in AI appropriate for our research , such as fuzzy logic , Bayesian network and rule-based . This research select rule-based since it is the most suitable approach fit with our needs.

Rule based is a methodology used in AI has the ability to perform the following functions according to [28]:

a. Learn from new situations

b. Apply solutions from old problems to new ones

c. Communicate about any topic, even if not familiar

d. Apply abstract reasoning to a range of specific problems

e. Understand how the world works in general

According to [28] Rule based systems usually designed to capture a domain expert's knowledge in propositional, qualitative form Separate the knowledge from the reasoning (inference) and from the specific facts. Allow heuristic reasoning and explanation.

A typical rule-based system has four basic components: [29] A list of rules or rule base, which is a specific type of knowledge base. An inference engine or semantic reasoner, which infers information or takes action based on the interaction of input and the rule base. Figure 2.10 shows the architecture of the rule base expert system.

Rule based expert systems according to [30] either forward chaining or back ward chaining. An example by Negnevitsky for forward chaining:

Y & D ' Z

X & B & E ' Y

A ' X

C ' L

L & M ' N

Database contains: A, B, C, D, E

In [N3;N4] author discussed a new web-based Programming Tutoring System based on student modeling process .The system guides the student into relevant information using AI abilities. This system is a good example for using rule based instructions in intelligent learning process.

Figure 2.10: Rule base architecture [31]

We are going to use forward changing in our research methodology same as the previous rules. Chapter 3 will discuss this matter in detail

3.2 Chapter summary

In this chapter we discussed various fields and models related directly to our research. The usage of these fields will be discussed in details in chapter 3. While chapter 3 will show how we used these models. As we discussed in the introduction our research should lead to design a new component for AVCM model. This part represents the student modeling process within the adaptive educational chat room .We discussed the architecture of AVCM and how we could integrate our results with it. Some methodologies are required to support our work are discussed too in this chapter, such as text and data mining. This field is considered the most important process in our work in addition to evaluation of time activity done by the student. Another fields underlined our work are needed to support our work such as regression analysis and rule based expert system.

'

Chapter III

Research Methodology and Evaluation

3.1 Methodology Introduction

This study is a hybrid study contains topics from several disciplines. We can find requirements from Education, web technology, computer science, and psychological field, Statistics as well. For this reason, having more than one methodology to perform this study is required strongly here. Theoretical and practical work will be joined together in this research.

3.2 Organization of the research

This study is composed of two components:

First: The theoretical component. This part will be built based on methodologies and theories already used by other systems in the field of chat rooms, student modeling, and domain modeling ,data mining , and time evaluation. These methodologies will help us to create the most suitable model to support the educational chat room in Open learning. Theoretical framework will rely mainly on the following concepts, which are already discussed in chapter Two:

1. Facilities offered by the virtual classroom should fit with the Open University. Such that the model should include the following facilities:

a. Lectures presentation

b. Online assessment (exams and assignments)

c. Online discussion tool (chat room)

2. The taxonomy of AVCM model

In reference to figure 2.1, by which we can find the framework of adaptive virtual classroom model. At the same time, we can discover the integration process with this system, which is missed in this model, which is the adaptive assessment and adaptive discussion. We will deal with this missing in the next chapter. Figures 2.2 and 2.3 will give us the main orientation of our adaptive chat room usage and advantages. The new architecture of AVCM should be as seen in the upgraded figure 3.1 shown below.

Figure 3.1: upgraded architecture of AVCM

3. Data and text mining

In reference to figure 2.7, this architecture will give use the conceptual framework of how to build our scope for chat text extraction and analysis. Text analysis will lead to text model which represents one main model in our work. Figure 2.7 is upgraded to be as shown in figure 3.2 to fit more with our work.

Figure 3.2: chat text mining in AVCM model

We will look for data mining in the chat to have the following situations and conditions:

1. The discussion topic usage in the chat

2. Number of keywords used in chat

3. Number of overall words used in the chat

4. Number of question terms used in chat session (What, Why, How').

5. Number of thanks words in the chat session (Thanks, good, excellent, understand, clear).

6. Number of rejection terms used in chat session (Sorry, not clear, not good, not understand).

All these points will be extracted using chat text mining for the two peers and analyzed .At the end of this research an equation for text modeling will be extracted.

4. Time evaluation

Reference to section 2.2.5. We are going to find the time usage by the students within the chat room. The parameters which are considered parts of time activities will be extracted and evaluated if they could be useful for the student modeling

5. Peer evaluation techniques

In reference to section 2.2.4 and figure 2.8.We are going to design a questionnaire to let each student evaluate his peer about the chat session. The evaluation process will lead to two outputs, first the score of the student and second the student characteristics evaluation.

Second: The practical component. This part will be used within the university itself and directly to the target students. This part enables us to find the most important parameters which we will use for time modeling, text modeling and peer assessment. These models we should consider in the system and also give us the result of our work.

This includes:

1. Gathering data from the target students, this data will provide us with following requirements:

2. Most important characteristics that should be embedded in the student modeling techniques, such as student preferences, interests, cognitive styles, background, age, etc.

3. Gathering data from Experts (teachers).

4. Hold two experiments and analyze the results

3.3 Results and Evaluation

So far, the Open University in Palestine, which is selected as the center of our data collections and experiments, uses high-tech virtual classrooms for its students in open education. However, these virtual classes deal with all students in the same way, same scientific level and the same level of testing. To obtain useful results we must apply our scientific methodology in the framework of this university, which depends entirely on open education, the students in this university have different characteristics we can summarize as the following:

1. Different ages: not all of them are in the range of 18 to 20, but we can find students in the range of 18 to 60 years old.

2. Different background: not all of them have the same computer and internet knowledge for example. In addition, some of them already lost much information they acquired during the school era, while others gain more knowledge in some field because of their field of work.

3. Different interests: In the same class we can find students with interests like finishing the study in the shortest period, while others they don't care about the period and they want to have more knowledge.

4. Different preferences: These students have different preferences for learning process; some of them for example like to learn in the day time while others in the night time. Many students lost the ability to read books and papers, so they need another way to learn for example video or animations or even experiments.

5. Different knowledge level: when the course starts, students started to follow the presentations, assignments, and exams. At one specific time, these students will reach a level of acquiring knowledge actually not equal. It's not fair to continue with all students with same race, and consequently some of them will be in good situations while others lose everything.

More characteristics are involved in this direction. During our study we need to determine exactly which are the most important characteristics in this university mostly affect the student learning process and use these characteristics to build the student model.

3.4 The process of the research

The following steps are included:

1. Design our project based on AVCM architecture and its components.

A. Invest the functions provided by AVCM described in Figure 2.1 as adaptive online presentation in the Open University to state the knowledge level about the concept to be discussed in the chat room.

B. Create customized user interface for each student according to his knowledge level that allows making the collaborative discussion between students and not concentrated the discussion towards the teacher. The system must take into account the capabilities of other students in the discussion room and their knowledge about relevant topic of discussion.

C. Find the optimal solution for the representation of knowledge acquired by the student through discussion with other students or with the teacher in order to update the student

D. The interface design [N1]as shown in figure 3.3 and figure 3.4 shows that the main components of the educational chat room are:

1. Current student

2. Concept name

3. Session number

4. Session time counter

5. Peer evaluation function

Figure 3.3: interface design for AVCM chat

Figure 3.4: AVCM Chat Interface

E. Add our research results to the interface of the chat room shown in figure 3.2.

2. Experiments

Two experiments will be held within the campus of the University for Three Objectives:

A. Extract the parameters which are needed to use for time and text modeling.

B. Design and implementation of the peer evaluation form.

C. Check the possibility and success of our model in open learning.

3. Regression analysis.

Regression analysis as described in section 2.2.6 is a statistical methodology used to find the relation among parameters. Experiments will give us many parameters from text and time analysis .We will use regression analysis to find the final equation for the following:

A. Text modeling

B. Time modeling

C. Final student modeling consists of text, time and peer evaluation modeling.

4. Design and construct the adaptive chat room interface according to the student modeling discussed in last point. The interface should give the peer knowledge about his peer status in the last session. The process relies on AI technologies and reasoning techniques offered by rule based expert systems.

5. Writing the results and finish the research .However, the work still needs many years to be a complete system.

Chapter Four

Student modeling in adaptive educational chat room

4.1 Time modeling adaptive educational chat room

As we are describing the student activity on the chat room to model this student. We need to be sure about two points: first if the student obeys the law and attend the chat tool and spent the complete required time. Second if the student was active during the chat session. To achieve these two points we need to calculate the session time and activity time.

This issue actually cuts to the heart of a fundamental problem with chat room analytics ' the notion of what is a 'student' and the assumptions that we use to determine this.

To understand how time on chat room works it is important to understand how this metric is calculated in your web analytics software. The Session time that a student spends on the chat room is calculated from the difference between login and log out time, while the activity time is calculated as the difference between the recorded time of his posts on the chat form not the time of login and log out.

In this way we can consider that leaving the chat room open will cause the activity average time to be greater and less and consequently indication of low activities by the student. For example the students posts first message at 1st second , and next post at 30th second , the average of these two posts is 15.5 while for other student the posts may be at second 2 , 5, 9.15 ,21 , and 28 .The average of second student is 13.3 . In this case we can be sure that second student has better activity than the second student because he has average activity time less than first student.

If the average for all students to the chat room was 3 minutes 20 seconds we could reasonably infer that student 1 where more interested in the chat room than student 2. Also as student 2 had a higher time on site than the overall average, we could say that these students were less interested than the general students to the site.

Activity time is not only the average time of posts .We need to be sure about the gap among posts time is fair. For example the student may still Idle long time before starting to participate heavily in the chat , or may be the participation is concentrated in some intervals and empty in other intervals . To solve this problem we can use two types of activity time, activity time participation and activity time average. Activity time participation (TR) is represents the rate of posts in the total time of the session while the activity time average represents the rate of pots time for the number of all posts. At the same time we need to consider another two variables in the chat session, the initial and end idle time. We need to calculate these two variables to check if the student activity is concentrated in some interval of time chat time or well distributed overall the chat time. In figure 4.4 we can see the flow chart of the time modeling process.

4.1.1 Time parameters extraction and evaluation

To check and evaluate the parameters required to model the time in chat session, an experiment was held in Palestine technical college ' Arroub. 18 students were given 45 minutes to discuss one concept in database course. 14 out of the 18 students started and completed the chat session while the others didn't take it serious and didn't make conversation .The concept was relationships and keys. All the students studied this concept within the classroom with an assistant professor who gave us and evaluation for each student from his perspective as an expert according to the student chat conversation. The students are from database and software department, second year and they are different in their knowledge level. The experiment is seeking for two main factors. The first factor was the effective time used by the student which we call it Time modeling, and the second factor is conversation done by the student and its value which we call it text modeling. The students used Gmail chat tool and we got the following results.

4.1.2 Time analysis results for experiment 1

We registered the posts intervals for each student during the chat session as seen in table 4.1 sample.

Table 4.1: time intervals registration sample

Number T1 T2 T3 T4 T5 T6 T7 T8

Adaptive 1 14 27 28 29.5

Adaptive 11 27.2 27.4 29 30

In table 4.2 we divided the time evaluation criteria according to time evaluation experts perspective into eight elements described in table 4.2:

Table 4.2: Suggested criteria for time modeling

Criteria Code Description

Time intervals average TIA Finding average of posts time

Time intervals median TIM Finding median of posts time

Time intervals count TIC Number of intervals

Activity Rate AR Student activity in active time

Total rate TR Student activity in complete time

IDLI IDLI Initial IDLE time

IDLE LDLE End IDLE time

Total IDL TIDL Complete IDLE time

As seen in table time, we considered only the IDLE time at the beginning of the session and at the end of the session, because it is highly complicated to find out the IDLE time within the session. But it's not completely ignored, the IDLE time is considered by other criteria like TIA and TIC by default.

After finishing the experiment, and registered the time intervals for all students as seen in the sample seen in table 4.1, we got the following results as seen in table 4.3.

Table 4.3: Experiment result

Number TIA TIM TIC AR TR New IDLI New IDLE New Total IDL

adaptive 1 41.7 44.0 21 47.73 46.67 1 1.5 2.5

adaptive 11 41.4 40.5 31 76.54 68.89 14.2 2.9 17.1

adaptive 3 42.3 42.1 44 104.51 97.78 12 1.2 13.2

adaptive 13 42.2 43.0 29 67.44 64.44 14.2 1.4 15.6

adaptive 5 37.9 37.3 55 147.45 122.22 0 1.8 1.8

adaptive 15 39.5 39.4 37 93.91 82.22 12 1.7 13.7

adaptive 6 44.1 42.2 41 97.16 91.11 20 0.5 20.5

adaptive 16 42.8 40.8 32 78.53 71.11 20.3 0.2 20.5

adaptive 7 39.0 38.6 52 134.89 115.56 0 0.5 0.5

adaptive 17 39.0 37.5 97 258.67 215.56 12 1 13

adaptive 8 40.7 38.0 53 139.47 117.78 1 0.6 1.6

adaptive 18 43.0 43.0 31 72.09 68.89 13.1 0.3 13.4

adaptive 9 39.7 39.9 111 278.20 246.67 13.2 3.1 16.3

adaptive 19 40.0 39.8 61 153.27 135.56 13 3.3 16.3

When the statistical analysis was performed, we found that the posts average and median time almost equals for all student as seen in figure 4.1. The difference between TIA and TIM is shown in table 4.4.

Table 4.4: Difference between TIA and TIM in the experiment

Number TIA TIM difference new

TIA-TIM

adaptive1 41.7 44 -2.33

adaptive11 41.4 40.5 0.89

adaptive3 42.3 42.1 0.16

adaptive13 42.2 43 -0.79

adaptive5 37.9 37.3 0.57

adaptive15 39.5 39.4 0.14

adaptive6 44.1 42.2 1.89

adaptive16 42.8 40.8 2.09

adaptive7 39 38.6 0.49

adaptive17 39 37.5 1.45

adaptive8 40.7 38 2.67

adaptive18 43 43 0.01

adaptive9 39.7 39.9 -0.19

adaptive19 40 39.8 0.21

average 0.363

round average 0.36

As seen from table 4.4, the average of the difference between median and average is 0.36. No significant difference between the two parameters and they could be used interchangeably. We can use the median time which can be sufficient for the required result.

Figure 4.1: Posts Average and median interval time

Activity rate and total rate time are two parameters should be calculated using average of median of the posts intervals. As we said before we can use the median to calculate the two new parameters. Activity rate is the rate of posts count to the posts intervals while total rate is the rate of posts count to the total time of the session. The two parameters statistically described in table 4.4 and in figure 4.2 we can see the difference between AR and TR values.

Table 4.5: The result of AR and TR

Number AR new TR new

adaptive 1 47.73 46.67

adaptive 11 76.54 68.89

adaptive 3 104.51 97.78

adaptive 13 67.44 64.44

adaptive 5 147.45 122.22

adaptive 15 93.91 82.22

adaptive 6 97.16 91.11

adaptive 16 78.53 71.11

adaptive 7 134.89 115.56

adaptive 17 258.67 215.56

adaptive 8 139.47 117.78

adaptive 18 72.09 68.89

adaptive 9 278.20 246.67

adaptive 19 153.27 135.56

Figure 4.2: Activity rate and Total rate

Area highlighted by red color represents total rate while the area heighted by blue represents active rate. The green color represents the difference between the results of the two value .It is reasonable to consider that mostly the two values have uniform difference which enables us to consider only one value. We are working with uncertain environment and we are looking for the most precise evaluation and not complete evaluation. To decide about parameter we should select it is required to calculate the standard deviation of the two parameters. TR has Stdev =57.49 while AR has Stdev =68.89. TR has les deviation than the AR which makes it more convenient than AR.

As a result shown from figure 4.2 shows that the two parameters have direct proportion. If the rate active increases the rate total also increases almost in uniform rate. As we see from the experiment result we may not be able to have percentage parameter when we use rate active and because the two parameters are completely related in the same direction we found that using one parameter is more accurate and enough for the evaluation process. We used total rate as one evaluation parameter in time modeling.

Total rate = posts count / complete session time ''. (4.1)

Back again to the other parameters, IDLI and IDLE .These two parameters are considered as wasted time and affect strongly the activity of the chat conversation. The total Rate we considered in equation (4.1) describes the activity of the student within the chat session time for all students equally, but for those students who don't use the chat time effectively usually have wasted time which we call it IDL time .For this reason it is reasonable to evaluate the time activity model from the total rate described in equation (4.1) in addition to the negative effect of IDL time on the chat activity. To do that we add TIDL to the complete session time as seen in equation (4.2):

Total activity rate = posts count / (complete session time + IDL)''. (4.2)

Equation (4.2) checks the negative behavior of the student that increasing the IDL time and consequently decreases the evaluation grade. The final result of the student's active time according to equation (4.2) will be as in the following table 4.6:

Table 4.6: Student total activity rate according to equation (4.2)

Number Rate Expert score

adaptive 1 44.2 55

adaptive 11 49.9 60

adaptive 3 75.6 75

adaptive 13 47.9 50

adaptive 5 117.5 112

adaptive 15 63 45

adaptive 6 62.6 40

adaptive 16 48.9 38

adaptive 7 114.3 90

adaptive 17 167.2 130

adaptive 8 113.7 80

adaptive 18 53.1 35

adaptive 9 181.1 140

adaptive 19 99.5 40

Regression analysis helps use understand how the typical value of the relationships among variables and gives us the equation of the time modeling .

Till now we have one variable which has uncertain value because of the changeable environment in which we calculate the value. In order to get more accurate of the total activity rate, the chat posts intervals are checked by time and education experts who have us an evaluation value for the student's activity in the chat room. This value is shown in table 4.6 as expert score and the relationship between the total activity rate and the expert evaluation is shown in figure 4.3.

Figure 4.3: Relationship between total activity rate and expert score

Two variables one dependent and one independent. We need to find the relationship between them by regression analysis as seen in table 4.6. The result was found as seen in equation (4.3):

Time model = 8.553 + 0.703 (Rate)'''''' (4.3)

When we use the full parameters equation (4.3) will be as the following equation (4.4)

Time model = 8.553 + 0.703 * posts count '''''. (4.4)

4.1.3 Time analysis results for Experiment 2

Another experiment was held to have more accurate result for time modeling. The new experiment was also applied on 14 students with new concept .The new concept is numerical systems in computer. The table 4.7 shows the result of this experiment.

Table 4.7: Time evaluation result in experiment 2

'

Rate Total IDL Posts count

Number

170.7 0.1 77 adaptive 1

158.7 1 73 adaptive 11

121.8 5.1 61 adaptive 3

163.3 9.5 89 adaptive 13

282.6 0.3 128 adaptive 5

157.9 12 90 adaptive 15

144.1 0.1 65 adaptive 6

157.7 10.8 88 adaptive 16

99.6 9.2 54 adaptive 7

180.7 9.8 99 adaptive 17

151.1 2 71 adaptive 8

171.2 2.3 81 adaptive 18

123.1 2.1 58 adaptive 9

174.1 2.1 82 adaptive 19

For the students in experiment 2 also we use expert evaluation as seen in table 4.8

Table 4.8: Student total activity rate according to equation (4.2) with expert score

Number Rate Expert Score

adaptive 1 170.7 150

adaptive 11 158.7 170

adaptive 3 121.8 130

adaptive 13 163.3 160

adaptive 5 282.6 200

adaptive 15 157.9 150

adaptive 6 144.1 112

adaptive 16 157.7 134

adaptive 7 99.6 100

adaptive 17 180.7 170

adaptive 8 151.1 160

adaptive 18 171.2 210

adaptive 9 123.1 190

adaptive 19 174.1 160

Regression equation (Experiment 2) = 85.073 + 0.445 (Total rate) '''. (4.5)

Regression equation (Experiment 1) = 8.553 + 0.703 (Total rate) ''' (4.6)

Now to take the best accurate result we can take the parameters average for both equations (4.5) and (4.6)

Time model 46.813 + 0.574 * (posts count / Complete session time + TIDL) '' (4.7)

The time model for the last two experiments is shown in table 4.9

Table 4.9: Time model after applied average time equation (4.7)

Number Rate ex1 Rate ex2 Time Model 1 Mark for time model 1 from 100 Time Model 2 Mark for time model 2 from 100

adaptive 1 44.2 170.7 75.5 50 144.8 69

adaptive 11 49.9 158.7 90.2 60 137.9 66

adaptive 3 75.6 121.8 74.3 49 116.7 56

adaptive 13 47.9 163.3 114.3 76 140.5 67

adaptive 5 117.5 282.6 83 55 209 100

adaptive 15 63 157.9 82.7 55 137.4 66

adaptive 6 62.6 144.1 74.9 50 129.5 62

adaptive 16 48.9 157.7 112.4 75 137.3 66

adaptive 7 114.3 99.6 142.8 95 104 50

adaptive 17 167.2 180.7 112.1 74 150.5 72

adaptive 8 113.7 151.1 77.3 51 133.5 64

adaptive 18 53.1 171.2 150.8 100 145.1 69

adaptive 9 181.1 123.1 103.9 69 117.5 56

adaptive 19 99.5 174.1 46.8 31 146.7 70

4.1.4 Time modeling algorithm

From the data collected and the experiment performed we can build our own algorithm which can be used to model the time within the chat room for each student alone. This algorithm is one the evaluation tool used for adaptive chat tool modeling, beside the text mining and peer evaluation. Figure 4.4 shows the full algorithm used for time modeling in chat room.

Figure 4.4: Full algorithm used for time modeling in chat room.

4.4.5 Time Model description

To describe the algorithm we need first to describe the acronyms used in the flow chart:

Table 4.10: description of acronyms used in the flow chart for time model

acronym Term Function

CT Chat topic Extraction of the discussion concept from the course domain model

TST Complete session time The total time pre-specified for the current session

CS Chat student The current student in the chat session

PC Post counter The number of posts done by the current student during the current session

I Intervals Sum of time posts time intervals during the current session

FPT First post time The time of first message posted by the student

EPT End post time The time of last message posted by the student

STE Session time evaluation The total time the student stayed logged in the chat session

TR Total rate The student participation in chat posts. Evaluate the number of posts in the total time of the session

AR Activity rate The rate of the sum of the time intervals of posts to the number of submitted posts

EIDL End Idle The Idle time at the end of the session

IIDL Initial Idle The Idle time at the beginning of the session

AM Activity model The model that describes the activity during the session

TM Time model The model that describes the time during the current session for the current student.

The time model in the chat session for the current student as extracted from the algorithm is:

CS, CT, STE, AM

While Activity model which is needed to model the time is:

TR, AR, EIDL, IIDL

The activity time model will describe and ensure that the student participates efficiently in the chat session and doesn't login and log out without be active during the session. Another two factors affects the activity model which is the time spent without submission of any posts either at the beginning of the chat session or at the end of the chat session .At the same time the total attendance time in the session is calculated by the difference between the login and logout time. The final time model is described using the activity model and session time evaluation.

Example: Let us consider the following data for one chat session.

Table 4.11: example for time model for one chat session

Factor Student 1` Student 2

CS X Y

CT BN BN

TST 60 60

EPT 55 50

FPT 4 10

Log in Time 2 0

Log out time 57 55

STE 55 55

PC 10 8

I 45 33

AR 10/60 = 0.17 = 17 8/60=0.13 = 13

TR 45/10 = 4.5 = 45 33/8 = 4.1=41

EIDL 5 10

IIDL 4 10

AT (0.17 , 4.5 , 5 , 4) (0.13 , 4.1 , 10 , 10)

Time model for the first student is:

(X , BN, 55 , 17 , 45 , 5 , 4)

Time model for the second student is:

(Y , BN , 55 , 13 , 41 , 10 , 10)

The evaluation last values affect the student model positively and negatively. The Idle entries are negative values while the other values are all positives .In this case we can extract the last student evaluation according to time modeling for the last two students as samples for the rest of the students participated in the same chat session. The evaluation is shown in the following table 4.12:

Table 4.12: example of evaluation time model in one chat session

model St - X St 'Y

STE 55 55

ATP 17 13

ATA 45 41

EIDL 5 10

IIDL 4 10

Evaluation 108 89

4.2 Text modeling in adaptive educational chat room

With the continuous using of chat rooms in one to one collaborative learning, it became essential that this chat should be modeled in order to evaluate the participants in this chat. It is not logic that we leave the conversation disoriented and out of control which encourages the students to deal with educational chat room as traditional one .The students will start to make general or private conversation which is not related directly or indirectly to the chat objectives. We are looking for factors that proof that the conversation is serious and related to the concept of discussion. Text mining is the best solution for our goal.

Text mining is generally the process of extracting interesting information and non-trivial knowledge from unstructured text. Text mining usually relies on artificial intelligence. Text mining is different from search is that the search is required for the user to know what he or she is looking for , while text mining is to discover information in a style that is not known in advance. Text mining should give knowledge as a result.

In our case we need to find solution for the text evaluation in the chat session conversation between two students .consequently we need to design an algorithm to solve this problem. Chat conversation is unstructured text .It is very difficult to analyze if we don't convert to structured text .Structured information is easier to search, manage, organize, share and to create reports on, for computers as well as people, hence the desire to give structure to unstructured information. This allowing computers and people to better manage the information, and allow known techniques and methods to be used.

Using text mining, instead of searching for words, we can search for linguistic word patterns, and this is therefore searching at a higher level.

The text in the chat conversation usually has many expressions; some of them can be used for the text modeling while the others are not used at all. Even the used expressions have different distance from the concept of discussion which is the core of our modeling process. These expressions are extracted from the experiment we conducted in PTCA.

Referring to the experiment held in Palestine technical college. 18 students was applied the experiment discussing the relations and keys in relational database system. From the experiment we find that the expressions used in the chat conversation are either useful or not useful .We divided the useful expressions eight clusters as seen in table 4.13.

Table 4.13: Chat text clustering

Category level Target

Main concept Level 1 Self

Related directly to the main concept Level 2 Self

Useful words Level 3 Self

Positive expressions Level 4 Both

Agreement expressions Level 5 Both

Enquiry expressions Level 6 Both

Respect expressions Level 7 Self

Negative Expressions Level 8 Peer

Unused Expressions Level 9 Both

Nine clusters are shown in table 4.13. We called them as levels in our study because they are closed or far from the main concept of the discussion in the chat session. The last level is unused cluster which contains the expression out of other clusters and not directly affect the text model in the chat session. We describe each level as in the following table 4.14:

Table 4.14: Levels description

Cluster Content

Level 1 Relation , keys , degree

Level 2 Table, one, many, primary, foreign, composite, database. information ,field , record , operation , process , insert , query , redundancy , merge , map , related , attribute , unique , redundancy , number , text , element , system , binary , empty , Null , triple , circular , access , SQL , compound . (and all other derivatives)

Level 3 Example, instance, suppose, declaration, describe, allowed, specified, See ,Look, Refer ,Moreover , However , Anyhow ,Despite , In spite , yet ,classify.( (and all other derivatives)

Level 4 Clear, Ok, good, excellent, understand, Thanks, yes, amnesty, correct, true, of course

Level 5 Not clear ,Bad , I don't know , wrong ,

Level 6 That is right, You are right, I think so too, agree with you, Exactly, Right, well, absolutely true.

Level 7 Can , How , what , when , where , who , which , Ask you , Give me , Describe me ,

Level 8 Hello , high , how are you , good morning , good day , good afternoon , good evening , have a nice time , gentle .

Level 9 All other words

The last 8 levels can describe the conversation from different perspectives which can give us the text evaluation and modeling ability. In order to model the text we need to cluster. The above mentioned levels as described in the following figure 4.5:

Figure 4.5: Chat text clustering

These parameters are all extracted from the experiment within Palestine technical college .The parameters are divided into three main categories, one for expressions used for student evaluation and the other for those expression which are not important for us and should be removed completely, such as is, are, you, he, for, and other not related directly or indirectly to the chat concept. The used parameters by its turn is divided into two categories because we need to have component based system .First category is user defined which is related directly or nearly to chat concept. This category should be defined for each concept by the instructor himself .These two parameters will be merged in one level during the modeling process. The second category is not related to the concept. It is related to expressions give us an indication about the chat activity and scenario. The second category is a component can be used for any concept which makes it more accurate than the first category because it can be updated frequently once we have new expressions could be used in this category. The last category counts the number of words used by each student within the chat session.

4.2.1 Chat text experiment analysis

The Experiment held in Palestine technical college on 18 students, 4 out of them failed to apply the experiment efficiently while the others applied it successfully. Table 4.15 describes results of the chat text for all the students according to the Parameters described above.

Level 1 to level 8 in addition to the total words used in the chat session is described in table 4.15. These values are extracted using database system through matching keywords in the database with the chat expressions and count the number of matching expressions for each level. As we described in the early stages of this dissertation, the system should database oriented and component based system. Figure 4.6 describes the main component of the system.

Figure 4.6: Proposed architecture of the system

Applying the experiment on the model described in figure 4.6 means that all the students chat text should be converted from the chat window into text file for easier manipulation.

The words in the text file should initially counted and store in a variable the it should be read sequentially, one word by one word .These words are considered as keyword .The search engine will look for the keyword and its derivatives in the database files designed for this propose .If found it means it is used keyword, otherwise it will be considered as not used keyword. Finally we will have a count for each parameter or level which can be used for text modeling in the chat.

The experiment result showed in table 4.15 gives us an indication about the usage of keywords from all levels for all students starting from words count until level 8. In our experiments the number of used words is less that used words .This is because the nature of the chat and the serious attitude of the students .For other students and other concepts may be this result will be different specially if the students understand well how the chat text is modeled and it concentrates on specific keywords and not all text used in the chat.

Table 4.15: Text analysis results

Number Total words Level 1 Level 2 Level 3 Level 4 Level 5 level 6 level 7 level 8 used words unused words

adaptive 1 191 12 36 13 8 1 0 8 1 79 112

adaptive 11 193 17 48 6 7 1 4 5 1 89 104

adaptive 3 349 15 73 18 5 2 8 5 3 129 220

adaptive 13 100 15 11 13 9 4 6 8 3 69 31

adaptive 5 442 21 89 17 4 0 1 2 2 136 306

adaptive 15 187 19 29 15 4 3 1 13 1 85 102

adaptive 6 116 8 28 9 3 1 4 9 1 63 53

adaptive 16 138 6 16 7 0 1 2 1 2 35 103

adaptive 7 166 4 34 5 6 2 1 5 0 57 109

adaptive 17 277 18 33 14 5 1 3 10 7 91 186

adaptive 8 120 9 12 8 3 2 10 18 3 65 55

adaptive 18 158 2 25 10 5 3 1 2 3 51 107

adaptive 9 271 8 25 5 3 8 0 3 1 53 218

adaptive 19 288 21 26 5 10 2 15 16 2 97 191

From the table 4.15 we can extract the following facts:

1. The average of used words is 40% from the total words, while the unused words are 60% for all students.

2. No significant relation between number of words and the other levels. It means that the student may talks a lot but without significant expressions. In the following table 4.16 we see the rate of levels to the number of words in chat session for each student.

Table 4.16: Levels weight from the total words in the chat

Number Total

words Level

1 Level

2 Level

3 Level

4 Level

5 Level

6 Level

7 Level

8

Adaptive 1 191 0.06 0.19 0.07 0.04 0.01 0.00 0.04 0.01

adaptive 11 193 0.09 0.25 0.03 0.04 0.01 0.02 0.03 0.01

adaptive 3 349 0.04 0.21 0.05 0.01 0.01 0.02 0.01 0.01

adaptive 13 100 0.15 0.11 0.13 0.09 0.04 0.06 0.08 0.03

adaptive 5 442 0.05 0.20 0.04 0.01 0.00 0.00 0.00 0.00

adaptive 15 187 0.10 0.16 0.08 0.02 0.02 0.01 0.07 0.01

adaptive 6 116 0.07 0.24 0.08 0.03 0.01 0.03 0.08 0.01

adaptive 16 138 0.04 0.12 0.05 0.00 0.01 0.01 0.01 0.01

adaptive 7 166 0.02 0.20 0.03 0.04 0.01 0.01 0.03 0.00

adaptive 17 277 0.06 0.12 0.05 0.02 0.00 0.01 0.04 0.03

adaptive 8 120 0.08 0.10 0.07 0.03 0.02 0.08 0.15 0.03

adaptive 18 158 0.01 0.16 0.06 0.03 0.02 0.01 0.01 0.02

adaptive 9 271 0.03 0.09 0.02 0.01 0.03 0.00 0.01 0.00

adaptive 19 288 0.07 0.09 0.02 0.03 0.01 0.05 0.06 0.01

For example in level 1 for adaptive 5, number of total words are 442 while the rate of levels to the number of words are:

Table 4.17: example for rate of levels to the number of words for adaptive 5

0.05 0.20 0.04 0.01 0.00 0.00 0.00 0.00

While for the student adaptive 13, number of used words are 100 while the rate of levels to the number of words are:

Table 4.18: example for rate of levels to the number of words for adaptive 13

0.15 0.11 0.13 0.09 0.04 0.06 0.08 0.03

In this case we can't consider the number of total words within the chat session as evaluation parameter in most cases.

3. The relationship between level 1 (the concept) and level 2 (the related) is shown in table 4.19 and figure 4.7.

Table 4.19: the values of level 1 and level 2

Level 1 level 2 Total

0.06 0.19 0.25

0.09 0.25 0.34

0.04 0.21 0.25

0.15 0.11 0.26

0.05 0.20 0.25

0.10 0.16 0.26

0.07 0.24 0.31

0.04 0.12 0.16

0.02 0.20 0.23

0.06 0.12 0.18

0.08 0.10 0.18

0.01 0.16 0.17

0.03 0.09 0.12

0.07 0.09 0.16

Figure 4.7: The relation between level 1 and level 2

From figure 4.7 we can see that level 1 and level 2 are mostly having the same rate. Considering them as one parameter is significant and more useful in our model. We can infer from figure 4.7 that sum of the two parameters mostly gives the same rate for both parameters. In this case we can consider the two parameters as concept parameter which equals concept + tangent as shown in table 4.19.

4.2.2 Text Parameters analysis

As we assumed before the number of words used in the chat by one student in one session is not a significant parameter to be used for student evaluation, but at the same time we can't ignore it completely. At the same time we need a base point to measure the distance of other parameters from it. It is logic that we can measure the other parameters relying on the number of words used in the session. The following table 4.19 shows the percentage of each parameter in reference to the number of words used in the chat session. By analyzing the other levels we didn't find any significant relationship among them .So we leave them as they are. The new table of parameters for all the students is shown in table 4.20.

Table 34.20: levels count for the students

Number Total

words Level

1 Level

2 Level

3 Level

4 level

5 level

6 Level

7

adaptive 1 191 48 13 8 1 0 8 1

adaptive 11 193 65 6 7 1 4 5 1

adaptive 3 349 88 18 5 2 8 5 3

adaptive 13 100 26 13 9 4 6 8 3

adaptive 5 442 110 17 4 0 1 2 2

adaptive 15 187 48 15 4 3 1 13 1

adaptive 6 116 36 9 3 1 4 9 1

adaptive 16 138 22 7 0 1 2 1 2

adaptive 7 166 38 5 6 2 1 5 0

adaptive 17 277 51 14 5 1 3 10 7

adaptive 8 120 21 8 3 2 10 18 3

adaptive 18 158 27 10 5 3 1 2 3

adaptive 9 271 33 5 3 8 0 3 1

adaptive 19 288 47 5 10 2 15 16 2

To understand the relation between each level and the total words used in chat we got the percentage rate for each level in reference to the total number of words and at the same time we calculated the rate for the same levels to used expressions only. We need a base to calculate the activity of the student in reference to this base. Logically we found that using the used expressions as the reference base is better than the total words of the chat since we are looking for the expressions used by the student and useful not the unuseful expressions.

We can take average for each level percentage for all students and take the percentage of each level to the average of used word as shown in table 4.21.

Table 4.21: Total words and used words as a reference base

Rate Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7 average

To the total conversation 0.2203 0.0484 0.024 0.0103 0.0187 0.035 0.01

Rounded decimal To the total conversation 22.03% 4.84% 2.4% 1.03% 1.87% 3.5% 1% 5.24%

To the used words 0.601 0.132 0.065 0.028 0.051 0.096 0.027

Rounded decimal places To the used words 60.1% 13.2% 6.5% 2.8% 5.1% 9.6% 2.7% 14.29%

Table 4.22 shows the result of this process for all students.

Table 4.22: levels rate to the total words used

Number Level1 Level 2 Level 3 Level 4 level 5 level 6 level 7

adaptive 1 25.1% 6.8% 4.2% 0.5% 0.0% 4.2% 0.5%

adaptive 11 33.7% 3.1% 3.6% 0.5% 2.1% 2.6% 0.5%

adaptive 3 25.2% 5.2% 1.4% 0.6% 2.3% 1.4% 0.9%

adaptive 13 26.0% 13.0% 9.0% 4.0% 6.0% 8.0% 3.0%

adaptive 5 24.9% 3.8% 0.9% 0.0% 0.2% 0.5% 0.5%

adaptive 15 25.7% 8.0% 2.1% 1.6% 0.5% 7.0% 0.5%

adaptive 6 31.0% 7.8% 2.6% 0.9% 3.4% 7.8% 0.9%

adaptive 16 15.9% 5.1% 0.0% 0.7% 1.4% 0.7% 1.4%

adaptive 7 22.9% 3.0% 3.6% 1.2% 0.6% 3.0% 0.0%

adaptive 17 18.4% 5.1% 1.8% 0.4% 1.1% 3.6% 2.5%

adaptive 8 17.5% 6.7% 2.5% 1.7% 8.3% 15.0% 2.5%

adaptive 18 17.1% 6.3% 3.2% 1.9% 0.6% 1.3% 1.9%

adaptive 9 12.2% 1.8% 1.1% 3.0% 0.0% 1.1% 0.4%

adaptive 19 16.3% 1.7% 3.5% 0.7% 5.2% 5.6% 0.7%

Level 1 to level 7, the rate of each level to the total used expressions is shown in table 4.22. Let us take an example the student adaptive 6 and adaptive 9. The first student have better rate than the second one because he has 31% while the second one has 12.2 %.Of course the increasing of the level 1 rate will decrease the rate of other levels but still significant if the decreasing rate occurs in last levels not the first levels . The experiment can't be the exact and final result for all levels. To be more precise we need to support our results with statistical analysis using regression calculation. Regression analysis need that we get an evaluation by experts for all students and level by level. To estimate the value of each parameter we submitted the chat text for all students to three experts in database systems to give us an evaluation for each student according to the target parameter in reference to the chat used words .The experts gave us the evaluation after reading the chat text. Their evaluation average is described in table 4.23.

Table 4.23: Experts evaluation for student's evaluation parameters in experiment 1

Number Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7

adaptive 1 26.8 9.0 4.6 0.6 0.0 4.6 0.3

adaptive 11 42.8 1.3 3.3 0.5 1.0 1.2 0.3

adaptive 3 29.1 6.1 1.0 0.7 2.0 1.0 0.7

adaptive 13 35.8 15.0 11.0 5.0 0.5 10.0 5.0

adaptive 5 33.4 4.0 1.0 0.0 0.0 0.3 0.7

adaptive 15 26.8 11.2 1.8 2.7 0.3 10.0 0.5

adaptive 6 37.5 10.6 1.2 1.2 2.9 11.6 0.4

adaptive 16 17.1 8.4 0.0 1.4 0.8 0.5 1.8

adaptive 7 25.8 1.7 4.0 2.2 0.5 2.7 0.0

adaptive 17 16.7 6.5 1.0 0.3 1.0 4.2 3.7

adaptive 8 10.2 8.4 1.5 2.9 12.8 20.0 3.0

adaptive 18 14.1 9.5 3.2 2.5 0.8 1.0 2.3

adaptive 9 13.7 0.9 1.0 5.0 0.0 0.5 0.3

adaptive 19 11.5 1.0 3.8 1.1 7.7 8.4 0.5

We consider the levels rate calculated in table 4.22 as the experiment evaluation for the parameters. So we remove the percentage and use the same table as seen in table 4.24 .We have two tables, table 4.24 contains the experiment result and table 4.23 contains the expert's evaluation. In order to get the proper value for each parameter we need to use the regression analysis which has the following form:

Y= b + aX

Table 4.24: experiment levels values

Number Level1 Level 2 Level 3 Level 4 level 5 level 6 level 7

adaptive 1 25.1 6.8 4.2 0.50 0.00 4.20 0.50

adaptive 11 33.7 3.1 3.6 0.50 2.10 2.60 0.50

adaptive 3 25.2 5.2 1.4 0.60 2.30 1.40 0.90

adaptive 13 26.0 13.0 9.0 4.00 6.00 8.00 3.00

adaptive 5 24.9 3.8 0.9 0.00 0.20 0.50 0.50

adaptive 15 25.7 8.0 2.1 1.60 0.50 7.00 0.50

adaptive 6 31.0 7.8 2.6 0.90 3.40 7.80 0.90

adaptive 16 15.9 5.1 0.0 0.70 1.40 0.70 1.40

adaptive 7 22.9 3.0 3.6 1.20 0.60 3.00 0.00

adaptive 17 18.4 5.1 1.8 0.40 1.10 3.60 2.50

adaptive 8 17.5 6.7 2.5 1.70 8.30 15.00 2.50

adaptive 18 17.1 6.3 3.20 1.90 0.60 1.30 1.90

adaptive 9 12.2 1.8 1.10 3.00 0.00 1.10 0.40

adaptive 19 16.3 1.7 3.50 0.70 5.20 5.60 0.70

4.2.3 Regression analysis for text parameters

In our study the value of Y is the expert evaluation while X is the experiment value. After performing the regression analysis on the last two tables we will get the following equations for each level as seen in table 4.24 b:

Table 4.24b : text levels equations

Level 1 equation: (-11.489+1.61x)

Level 2 equation: (-1.127+1.413x)

Level 3 equation: (-0.8+1.256x)

Level 4 equation: (0.074+1.416x)

Level 5 equation (-0.491+1.173x)

Level 6 equation: (-0.927+1.44x)

Level 7 equation: (-0.447+1.59x)

Equations of levels

5 Text Model =

0.601*level 1

+0.132*level 2

+ 0.065*level 3

+ 0.028* level 4

+0.051*Level 5

+0.096*level 6

+0.027*level 7

Text Model =

0.601*(-11.489 +1.61 *level 1)

+0.132* (-1.127 +1.413*level 2)

+ 0.065*(-0.8 +1.256*level 3)

+ 0.028* (0.074 +1.416 *level 4) ''''. (4.8)

+0.051*(-0.491 +1.173- *level 5)

+ 0.096* (-0.927 +1.44 *level 6)

+0.027*(-0.447 +1.59 * level 7)

Simplifying this equation we get the final Text model equation as the following:

To have the final parameters we describe the symbols in table 4.25:

Table 4.25: declaration of parameters

TW Total used words

Rlevel1 Rate of level 1 Level1 /used words

Rlevel2 Rate of level 2 Level2 / used words

Rlevel3 Rate of level 3 Level3 / used words

Rlevel4 Rate of level 4 Level4 / used words

Rlevel5 Rate of level 5 Level5 / used words

Rlevel6 Rate of level 6 Level6 / used words

Rlevel7 Rate of level 7 Level7 / used words

Basing on equation 4.8 and table 3.25we get the final equation of the text model as seen in equation 4.9.

Text Model =

-7.71 + (0.97 * Rlevel1) + (0.19 * Rlevel2) + (0.08 * Rlevel3)

+ (0.04 * Rlevel4) + (0.06* Rlevel5) + (0.14 * Rlevel6) + (0.04 * Rlevel7) ''.. (4.9)

4.2.4 Experiment number 2 for text analysis

To have more accurate results for the text model we conducted another experiment on new 14 students and with new concept. The new concept is numerical methods in computer system. We got the following results as seen in table 4.26.

Table 4.26: Experiment 2 Text analysis result

Number total words level 1 level 2 level 3 level 4 Level 5 level 6 level 7

adaptive 1 232 53 3 20 6 0 33 2

adaptive 11 353 118 22 14 5 3 16 3

adaptive 3 187 38 0 3 5 0 10 2

adaptive 13 246 69 2 7 10 1 5 2

adaptive 5 600 181 2 14 14 2 22 3

adaptive 15 229 22 3 11 4 3 23 2

adaptive 6 181 40 3 9 2 5 33 2

adaptive 16 383 110 5 6 7 2 14 3

adaptive 7 196 52 0 16 4 4 28 2

adaptive 17 258 57 3 10 8 4 5 2

adaptive 8 209 45 6 18 5 2 28 1

adaptive 18 303 93 4 11 5 1 3 2

adaptive 9 227 67 16 10 5 1 32 2

adaptive 19 186 43 6 12 20 3 5 3

We can take average for each level percentage for all students and take the percentage of each level to the average of used word as shown in table 4.27.

Table 4.27: Total words and used words as a reference base for experiment 2

Rate level 1 level 2 level 3 level 4 level 5 level 6 level 7 average

To the total conversation 0.2607 0.0198 0.0425 0.0264 0.0082 0.0678 0.0082

rounded decimal To the total conversation 26.07% 1.98% 4.25% 2.64% 0.82% 6.78% 0.82% 6.19%

To the used words 0.601 0.046 0.098 0.061 0.019 0.156 0.019 6.19%

Rounded decimal places To the used words 60.1% 4.6% 9.8% 6.1% 1.9% 15.6% 1.9%

Following the same methodology used in the last experiment by dividing the number of each level on the total words used in the experiment to extract the actual weight of each level, we got the following table 4.28.

Table 4.28: Levels rate in experiment 2

Number Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7

adaptive 1 22.8 1.3 8.6 2.6 0.0 14.2 0.9

adaptive 11 33.4 6.2 4.0 1.4 0.8 4.5 0.8

adaptive 3 20.3 0.0 1.6 2.7 0.0 5.3 1.1

adaptive 13 28.0 0.8 2.8 4.1 0.4 2.0 0.8

adaptive 5 30.2 0.3 2.3 2.3 0.3 3.7 0.5

adaptive 15 9.6 1.3 4.8 1.7 1.3 10.0 0.9

adaptive 6 22.1 1.7 5.0 1.1 2.8 18.2 1.1

adaptive 16 28.7 1.3 1.6 1.8 0.5 3.7 0.8

adaptive 7 26.5 0.0 8.2 2.0 2.0 14.3 1.0

adaptive 17 22.1 1.2 3.9 3.1 1.6 1.9 0.8

adaptive 8 21.5 2.9 8.6 2.4 1.0 13.4 0.5

adaptive 18 30.7 1.3 3.6 1.7 0.3 1.0 0.7

adaptive 9 29.5 7.0 4.4 2.2 0.4 14.1 0.9

adaptive 19 23.1 3.2 6.5 10.8 1.6 2.7 1.6

Again we need to the use the experts evaluation for each students for all levels in order to get the regression analysis result. The expert's evaluation can be shown in table 4.29.

Table 4.29: Experts evaluation for student's evaluation parameters in experiment 2

Number Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7

adaptive 1 25 2 15 3 0.0 20 2

adaptive 11 39 10 9 2 1 22 2

adaptive 3 19 0.0 0.5 2 0.0 12 3

adaptive 13 10 2 2 5 2 7 1.6

adaptive 5 50 1 7 6 2.6 22 3.6

adaptive 15 9 5 6 1 5 22 3

adaptive 6 17 3 5.0 1 6.2 35 3

adaptive 16 34 3 1 2 3 19 5

adaptive 7 25 0.0 11 1 6 35 3

adaptive 17 10 2 4 3 5 8 2.5

adaptive 8 11 5 13 1 1.5 29 2

adaptive 18 28 3 8 1 1 3 2

adaptive 9 23 13 4.4 1 1 19 1.5

adaptive 19 10 4 6.5 15 2 5 3

Using regression analysis for the results of experiment 2 we got the following text model:

1. Text Model =

0.601*level 1

+0.046*level 2

+ 0.098*level 3

+ 0.061* level 4

+0.019*Level 5

+0.156*level 6

+0.019*level 7

Text Model =

0.601*(-12.595+1.395*level 1)

+0.046* (0.431+1.648 *level 2)

+ 0.098*(-0.487+1.506*level 3)

+ 0.061* (-1.045+1.469*level 4)

+0.019*(0.581+2.167*level 5)

+ 0.156* (7.408+1.415*level 6)

+0.019*(2.311+0.39* level 7)

By taking the average of the two experiments and referring to table 4.25 we get the final form of the text model as the following:

Text Model average = -6.91

+ (0.9 * Rlevel 1)

+ (0.14 * Rlevel 2)

+ (0.11 * Rlevel 3)

+ (0.06 * Rlevel 4)

+ (0.06* Rlevel 5)

+ (0.18 * Rlevel 6)

+ (0.02 * Rlevel 7)

Using this equation we can calculate the text model for the students in the two experiments as the following:

Relying on Text and time model extracted from experiment 1 and 2. We need to have the score which is usually calculated out of 100. Since text and time model don't give a score actually, so we can take a base to convert the model to a score. It is suitable to consider the maximum text and time model as 100 scores and make a curve to extract the rest of the scores according to these marks. Table 4.30 shows the time and text score according to this procedure.

Peer evaluation is another model used in our experiment. This model enables the student to evaluate his partner in the chat session by filling in a questionnaire. The score should be out of 100. Now we have three input score and we need to have one output score for each student. The weight used for each input need to be calculated statistically by multiple parameter regression analysis technique. To perform that we need again the expert to give us general evaluation for each student according to his impression by reading the text written by the student individually. Table 4.30 and table 4.31 shows the final results of the 4 required parameters to be used in regression analysis which will give us the final equation for student modeling in educational chat room.

Table 4.30: Text and time score extracted from text and time model

Table 4.30: student model parameters in experiment 1

Number Time score Text score Peer evaluation Expert score

adaptive 1 50 72 68 69

adaptive 11 60 100 80 80

adaptive 3 49 69 65 70

adaptive 13 76 86 84 78

adaptive 5 55 65 82 70

adaptive 15 55 76 79 75

adaptive 6 50 96 88 85

adaptive 16 75 34 50 58

adaptive 7 95 61 69 71

adaptive 17 74 45 64 54

adaptive 8 51 53 66 62

adaptive 18 100 41 60 45

adaptive 9 69 19 80 43

adaptive 19 31 72 55 49

Table 4.31: student model parameters in experiment 2

Number Time score Text score Peer evaluation Expert score

adaptive 1 69 68 77 75

adaptive 11 66 100 82 86

adaptive 3 56 50 66 58

adaptive 13 67 76 74 70

adaptive 5 100 84 88 86

adaptive 15 66 17 34 40

adaptive 6 62 67 77 72

adaptive 16 66 79 75 74

adaptive 7 50 81 78 76

adaptive 17 72 56 66 68

adaptive 8 64 64 62 67

adaptive 18 69 85 88 80

adaptive 9 56 93 80 84

adaptive 19 70 64 67 65

Applying the multiple parameters regression analysis on table 4.30 and 4.31 we get the following equations for the two experiments as the following:

Score1(exp1) = 10.8058 + 0.0947167 *Time + 0.446414 *Text + 0.279352 *Peer

Score2 (exp2) = 15.325 + 0.128307 *Time + 0.401622 *Text + 0.267794 *Peer

Score average = 13.07 + (0.11 * Time Model) +(0.42* Text Model) + (0.27* Peer Model)

Time model 46.813 + 0.574 * (posts count / Complete session time + TIDL)

Text Model = -6.91

+ (0.9 * Rlevel 1)

+ (0.14 * Rlevel 2)

+ (0.11 * Rlevel 3)

+ (0.06 * Rlevel 4)

+ (0.06* Rlevel 5)

+ (0.18 * Rlevel 6)

+ (0.02 * Rlevel 7)

Peer Model = Questionnaire score

Complete student model =

2.29

+ (0.25*posts count / (Complete session time / TIDL))

+( 0.32 * Rlevel1 + 0.05 * Rlevel2 + 0.04 * Rlevel3 + 0.02 * Rlevel4 + 0.02 *

Rlevel5 + 0.06 * Relevel 6 +0.01 * Rlevel 7)

+0.3 * Peer score

Where Rleveli = Leveli/number of words

Applying the last equation on the results of experiment 1 and 2 we got the final student model for both experiments as shown in table 4.32 and 4.33.

Table 4.32: Student model for experiment 1

Number Time score Text score Peer evaluation final

score

adaptive 1 50 72 68 67

adaptive 11 60 100 80 83

adaptive 3 49 69 65 65

adaptive 13 76 86 84 80

adaptive 5 55 65 82 69

adaptive 15 55 76 79 72

adaptive 6 50 96 88 83

adaptive 16 75 34 50 49

adaptive 7 95 61 69 68

adaptive 17 74 45 64 57

adaptive 8 51 53 66 59

adaptive 18 100 41 60 57

adaptive 9 69 19 80 50

adaptive 19 31 72 55 62

Table 4.33: Student model for experiment 2

Number Time score Text score Peer evaluation final score

adaptive 1 69 68 77 70

adaptive 11 66 100 82 84

adaptive 3 56 50 66 58

adaptive 13 67 76 74 72

adaptive 5 100 84 88 83

adaptive 15 66 17 34 69

adaptive 6 62 67 77 69

adaptive 16 66 79 75 74

adaptive 7 50 81 78 74

adaptive 17 72 56 66 62

adaptive 8 64 64 62 64

adaptive 18 69 85 88 80

adaptive 9 56 93 80 80

adaptive 19 70 64 67 66

4.3 Peer evaluation process

Back to the peer evaluation model which we used in the last sections. This model represents a pop up questionnaire have three main sections with 27 multiple choice questions. Each question has weight 3.7 as a score. The answers will be as shown in table 4.33.a

Table 4.33.a: Peer evaluation questionnaire criteria

choice Strong agree Agree nothing reject Strong reject

Questionnaire score 5 4 3 2 1

Real score 3.7 2.96 2.22 1.48 0.74

QiS= Real score

Peer evaluation model = Sum(QiS)

The sections of the questionnaire have the following objectives:

1) Student Knowledge(PK)

2) Student text activity (PTXA)

3) Time activity (PTIA)

For each section the student should answer nine questions related to the objective of the section. Finally a score will be extracted according to the student answers. Next section will discuss this matter in more detail.

4.4 Student Model in Chat Interface

This model aims to create adaptive chat interface which helps the student to know about the status of his peer. The student can understand if his peer needs help, gives help or both. The process depends on inferring the status depending on the data about the student activity within the last chat session. To do that we depend mainly on the last peer evaluation process with the data gathered about the student himself in the last session (text and time model).The status of the student has three sections , knowledge , text activity and time activity .For that reason the peer evaluation process is divided into these three sections :

(PK, PTXA and PTIA) as the following:

1) Student Knowledge(PK) which consists of

' Main concept

' Related concept

' Useful words

2) Student text activity (PTXA)

' Positive reaction

' Agreement reaction

' Enquiry reaction

3) Time activity (PTIA)

' Total rate

' Activity rate

' TIDL

This is about the peer evaluation sections, which give three scores extracted from these three sections .About the knowledge, text and time model we also have three sections for each one matching with sections and entries of the peer evaluation entries. The parameters given by theses models are extracted through the student activity within the last chat session. For the knowledge, time and text activities we also have three parameters for each one of them as shown in figure 4.8 and table 4.34.

Figure 4.8: Parameters structure

At the end of the process we will have the adaptive chat interface as seen in figure 4.9.

Figure 4.9: adaptive chat interface according to the reasoning process

Table 4.34 shows the structure of the parameters used in the reasoning process. The reasoning process is a standard process used for the various sections in the same way as shown in the next tables.

Table 4.34: parameters structure used in the reasoning process

Category P1 P2 P3 P4

Knowledge PK M L S

description Peer K Main Concept Related concept Useful words

Text activity PTXA PO AG EN

description Peer text Positive Agreement enquiry

Time activity PTIA AR TR TIDL

Description Peer time Activity rate Total rate IDL time

Status Last OS KN TXA TIA

Final status Last overall session status Knowledge Text activity` Time activity

4.5.1 Reasoning process

The reasoning process finds an indication about the student status in the last session. This Indication will be represented using light traffic form. Table 4.35 shows the description of this indication.

Table 4.35: Indications description

Indication Status

Green Student can give help

Yellow Student can give and needs help

Red Student needs help

The problem now is how to know which student should be given a color as indication .In order to do that we need to refer to our last two experiments and take the final score column sorted in descend way as seen the following table 4.36.

Table 4.36: final score in descend order

number Final score Final score

1 83 84

2 83 83

3 80 80

4 72 80

5 69 74

6 68 74

7 67 72

8 65 70

9 62 69

10 59 69

11 57 66

12 57 64

13 50 62

14 49 58

We divide the final score into four sets, for each set we get the median of the scores within this set as seen in table 4.37.We considered 100 as maximum score and 35 as minimum score.

Table 4.37: sets of final scores and their median

Set scores median

Set 1 100 100

Set 2 83 84

83 83

80 80

72 80

69 74

68 74

67 72

77

Set 3

65 70

62 69

59 69

57 66

57 64

50 62

49 58

62

Set 4 35 35

Basing on table 4.37 we can find the indications for the parameters as described in the following facts:

Fact 1: P = Parameter

Fact 2: I=indication

Fact 3: P >= 77, I = Green (G)

Fact 4: P<77 and P>=62, I=Yellow (Y)

Fact 5: P < 62, I = Red (R)

These facts are a base can be used for all parameters used in this reasoning process either for knowledge state, text activity or time activity. As we discussed before we have four different parameters (P1, P2, P3, and P4), we consider all possible combinations for all parameters as shown in table 4.38. The center of the period is needed to find the nearest indication color .The period center is calculated as in the following equation:

Center = (Max + Min) / 2

And then the center of the periods will be as:

G center = (100+77)/2 = 88.5

Y center = (62+77)/2 = 69.5

R center = (35+62)/2 = 48.5

Table 4.38: parameters combinations and facts

Table 4.38 shows the possible combinations of the set described above, for each parameter we take center and indication for the level. For example the green color has the center as 88.5, yellow has the center as 69.5 and red has the center 48.5. By taking the average of these centers we get a new center represents the result of the four parameters average .Since we used three constant centers and consequently we need to use them as a base for our work we should take the nearest center to the center average and also have the indication for that center as final result for this part. If we have two equal distances with two centers we get the upper center as seen in table 4.35.

1. Rules:

GetStatus (parameter x)

Begin

If x >= 77 then color = Green (G)

If (x<77 and x>=62) then color =Yellow (Y)

If x < 62 then color = Red (R)

Return color

End

From table 4.38 we can write the rules in an algorithm form for the general reasoning process as shown bellow.

Procedure adaptivereasoning (P1,P2,P3,P4)

Begin

Input (P1, P2, P3, P4) values

GetStatus (P1)

GetStatus (P2)

GetStatus (P3)

GetStatus (P4)

Indication I

I= white

1. IF P1= G and P2= G and P3= G and P4= G then I= G

2. IF P1= G and P2= Y and P3= G and P4= G then I= G

3. IF P1= G and P2= R and P3= G and P4= G then I= Y

4. IF P1= G and P2= G and P3= Y and P4= G then I= G

5. IF P1= G and P2= Y and P3= Y and P4= G then I= G

6. IF P1= G and P2= R and P3= Y and P4= G then I= Y

7. IF P1= G and P2= G and P3= R and P4= G then I= Y

8. IF P1= G and P2= Y and P3= R and P4= G then I= Y

9. IF P1= G and P2= R and P3= R and P4= G then I= Y

10. IF P1= G and P2= G and P3= G and P4= Y then I= G

11. IF P1= G and P2= Y and P3= G and P4= Y then I= G

12. IF P1= G and P2= R and P3= G and P4= Y then I= Y

13. IF P1= G and P2= G and P3= Y and P4= Y then I= G

14. IF P1= G and P2= Y and P3= Y and P4= Y then I= Y

15. IF P1= G and P2= R and P3= Y and P4= Y then I= Y

16. IF P1= G and P2= G and P3= R and P4= Y then I= Y

17. IF P1= G and P2= Y and P3= R and P4= Y then I= Y

18. IF P1= G and P2= R and P3= R and P4= Y then I= Y

19. IF P1= G and P2= G and P3= G and P4= R then I= Y

20. IF P1= G and P2= Y and P3= G and P4= R then I= Y

21. IF P1= G and P2= R and P3= G and P4= R then I= Y

22. IF P1= G and P2= G and P3= Y and P4= R then I= Y

23. IF P1= G and P2= Y and P3= Y and P4= R then I= Y

24. IF P1= G and P2= R and P3= Y and P4= R then I= Y

25. IF P1= G and P2= G and P3= R and P4= R then I= Y

26. IF P1= G and P2= Y and P3= R and P4= R then I= Y

27. IF P1= G and P2= R and P3= R and P4= R then I= R

28. IF P1= Y and P2= G and P3= G and P4= G then I= G

29. IF P1= Y and P2= Y and P3= G and P4= G then I= G

30. IF P1= Y and P2= R and P3= G and P4= G then I= Y

31. IF P1= Y and P2= G and P3= Y and P4= G then I= G

32. IF P1= Y and P2= Y and P3= Y and P4= G then I= Y

33. IF P1= Y and P2= R and P3= Y and P4= G then I= Y

34. IF P1= Y and P2= G and P3= R and P4= G then I= Y

35. IF P1= Y and P2= Y and P3= R and P4= G then I= Y

36. IF P1= Y and P2= R and P3= R and P4= G then I= Y

37. IF P1= Y and P2= G and P3= G and P4= Y then I= G

38. IF P1= Y and P2= Y and P3= G and P4= Y then I= Y

39. IF P1= Y and P2= R and P3= G and P4= Y then I= Y

40. IF P1= Y and P2= G and P3= Y and P4= Y then I= Y

41. IF P1= Y and P2= Y and P3= Y and P4= Y then I= Y

42. IF P1= Y and P2= R and P3= Y and P4= Y then I= Y

43. IF P1= Y and P2= G and P3= R and P4= Y then I= Y

44. IF P1= Y and P2= Y and P3= R and P4= Y then I= Y

45. IF P1= Y and P2= R and P3= R and P4= Y then I= Y

46. IF P1= Y and P2= G and P3= G and P4= R then I= Y

47. IF P1= Y and P2= Y and P3= G and P4= R then I= Y

48. IF P1= Y and P2= R and P3= G and P4= R then I= Y

49. IF P1= Y and P2= G and P3= Y and P4= R then I= Y

50. IF P1= Y and P2= Y and P3= Y and P4= R then I= Y

51. IF P1= Y and P2= R and P3= Y and P4= R then I= Y

52. IF P1= Y and P2= G and P3= R and P4= R then I= Y

53. IF P1= Y and P2= Y and P3= R and P4= R then I= Y

54. IF P1= Y and P2= R and P3= R and P4= R then I= R

55. IF P1= R and P2= G and P3= G and P4= G then I= Y

56. IF P1= R and P2= Y and P3= G and P4= G then I= Y

57. IF P1= R and P2= R and P3= G and P4= G then I= Y

58. IF P1= R and P2= G and P3= Y and P4= G then I= Y

59. IF P1= R and P2= Y and P3= Y and P4= G then I= Y

60. IF P1= R and P2= R and P3= Y and P4= G then I= Y

61. IF P1= R and P2= G and P3= R and P4= G then I= Y

62. IF P1= R and P2= Y and P3= R and P4= G then I= Y

63. IF P1= R and P2= R and P3= R and P4= G then I= R

64. IF P1= R and P2= G and P3= G and P4= Y then I= Y

65. IF P1= R and P2= Y and P3= G and P4= Y then I= Y

66. IF P1= R and P2= R and P3= G and P4= Y then I= Y

67. IF P1= R and P2= G and P3= Y and P4= Y then I= Y

68. IF P1= R and P2= Y and P3= Y and P4= Y then I= Y

69. IF P1= R and P2= R and P3= Y and P4= Y then I= Y

70. IF P1= R and P2= G and P3= R and P4= Y then I= Y

71. IF P1= R and P2= Y and P3= R and P4= Y then I= Y

72. IF P1= R and P2= R and P3= R and P4= Y then I= R

73. IF P1= R and P2= G and P3= G and P4= R then I= Y

74. IF P1= R and P2= Y and P3= G and P4= R then I= Y

75. IF P1= R and P2= R and P3= G and P4= R then I= R

76. IF P1= R and P2= G and P3= Y and P4= R then I= Y

77. IF P1= R and P2= Y and P3= Y and P4= R then I= Y

78. IF P1= R and P2= R and P3= Y and P4= R then I= R

79. IF P1= R and P2= G and P3= R and P4= R then I= R

80. IF P1= R and P2= Y and P3= R and P4= R then I= R

81. IF P1= R and P2= R and P3= R and P4= R then I= R

Return I

end

2. Sections reasoning process

a. Knowledge state reasoning

We can consider the following facts from the last table:

Knowledge score in peer evaluation is PK

Main concept score in Text model is M

Related concept score in Text model is L

Useful word expression in Text model is S

Indication status is I

(PK, M, L, S, I) is a set represents chat session indicators.

We call the last procedure as bellow

Knowledge = Adaptivereasoning (PK, M, L, S)

b. Text activity reasoning

In text activity the facts are derived from the data given by the student behavior within the chat session. The text activity level of the student is determined by both the score extracted from peer evaluation and the three parameters (Positive expressions, Agreement expressions and enquiry expressions) the student used in the chat session. The section of text activity in peer evaluation model has the value PTXA, the text activity in text model has the value (PO: positive expressions), (AG: agreement expressions) and (EN: the enquiry expressions). The set is (PTXA, PO, AG, EN) .Each parameter in the previous set has three different levels. We follow same as last algorithm with the substitution of the parameters as the following:

Text = Adaptivereasoning (PTXA, PO, AG, EN)

c. Time activity reasoning

In time activity the facts are derived from the data given by the student behavior within the chat session. The time activity level of the student is determined by both the score extracted from peer evaluation and the three parameters (Total rate, Activity rate and Total IDL the student used in the chat session. The section of time activity in peer evaluation model has the value PTIA, the text activity in text model has the value (TR: total rate), (AR: Activity rate) and (TIDL: Total IDL). The set is (PTIA, TR, AR, and TIDL) .Each parameter in the previous set has three different levels. We follow same as last algorithm with the substitution of the parameters as the following:

Time = Adaptivereasoning (PTIA, AR, TR, TIDL)

The rules and facts used for text activity and time activity are the same as used for knowledge status. The reader can refer to them to review these rules.

At the end of the last three procedures we will get a new set of the three parameters, each one has three possible values or indications as seen in table 4.39:

Table 4.39: The new three parameters

Knowledge Text activity Time activity

G G G

Y Y Y

R R R

These parameters will describe the status of the student in the last chat session served for his current peer in the current chat.

3. Example:

To describe the last three procedures lets have the following example described in table 4.40:

Table 4.40: example of three parameters

PK S L M

88 73 90 73

Facts G G G Y Rule 2 G

PTXA PO AG AN

76 46 78 66

Facts G R G Y Rule 12 Y

PTIA TR AR TIDL

45 55 62 41

Facts R Y Y R Rule 77 Y

The status of the student will be as the set (G, Y, Y).

4. Student interface status modeling:

The final status of the student is an incremental modeling of the student. The current session will have 4 indications, knowledge status, text activity status, time activity status and the overall activity status. The overall activity status will take its value from the last session and inferring its value from the other three parameters. The overall status is denoted as OS .The result of OS will be given by calling the procedure as:

Status = Adaptivereasoning (Last OS, Knowledge, Text, Time)

The last example which had the status set as (G, Y, Y) will have the overall status if we consider for example the last OS as Y , the result according to rule 40 is .

Last OS Knowledge Text Time status

Y G Y Y Y

Chapter 5

Discussion and Conclusions

5.1 Introduction

This final chapter revisits the main aims and objectives of the work as stated in Chapter 1 and discusses the findings of the user trials conducted in the course of this research and the resulting implications for the negative results shown by the work. There is also some deliberation over the limitations of the work and the difficulties that are encountered when conducting educational research. Lastly, there are some suggestions of how the work could progress further.

5.2 Research objectives re-visited

The research objectives stated originally in Section 1.3 were:

1. New model of adaptive chat tool to be integrated with adaptive educational systems.

2. Modeling the student within this chat tool in order to reflect the knowledge level acquired with the chat session into the overall student profile through :

a. Peer evaluation model.

b. Chat context analysis model

c. Time spent in the chat model

3. Having adaptive chat tool interface based on the student modeling process

The first objective has been achieved by an extensive literature review for AVCM model and the possibility of integrating our model to AVCM model. Figure 3.1 shows the integration process and the possibility of achieving this objective.

The second objective have been achieved through the design and development of the modeling process discussed in chapter 4. Three models are discussed related to the context analysis , time analysis and peer evaluation which all represents the student modeling process within the chat room .Sections 4.1 , 4.2 and 4.3 discussed the procedures applied to have the three models . All these models equations are extracted through two experiments and with the help of experts 2.

The third objective was discussed and solved in section 4.4.The chat room interface -helps the student to understand the characteristics of his peer in the previous sessions. This may help the student to conduct his chat with his peer in better way.

5.3 Contributions to the field

As specified in Section 1.4, there was one major contribution arising from this research and one minor contribution. The major contribution of this work relates to the experimental approach and data analysis, which states that student modeling in educational chat room is possible, which has been under discussion always among researchers .Student modeling within the educational chat tool make the chat room more professional, controllable and evaluable. This made it oriented and the student will behave in serious way within it .Having a score for the students extracted from his activities in the chat session will be another input for the student score for this concept. This score will be added to the previous scores extracted during the virtual classroom e-testing process.

The minor contribution was the adaptive interface created relying on the activities done by the student during the sessions. The student will be able to know more about his peers so he can concentrate on the aspects more than other aspects according to the chat adaptive interface .When the student sees that the knowledge level tag is red for his peer, it means that he should rely on himself to conduct the discussion about the concept while he should rely on his peer if the color is green. The same thing will be done for text and time usage by the peer.

The current study that we made relied on three tracks. The first was through conducting extensive studies in adaptive educational systems available and get the ideas that we need in this study as we could know the weak points in such systems, then we have designed a model to serve the purposes that we have identified to achieve the goals of this research, therefore we created a prototype model based on the application of the concepts that we presented in the research. We have published one scientific paper confirm the validity of the methodology used in the education process. That was not enough, but we had to know the impact that holds this system to students through an experiment on a group of students to see this effect, as we studied the attitudes towards this adaptive style of education. The results that we got are relatively acceptable and showed us many positive and negative points in this model.

5.4 The model characteristics

5.4.1 Advantages

In addition to the main functions required by the system , providing adaptive ,controllable and evaluable educational chat tool model ,the model is component based , this gives the ability to add and change components without affecting the whole system .The model also is data independent , this gives the ability to use it for any course , only we need to change the entries of the main concept expressions in addition to the expressions related to the main concept .Other parameters for text and time models are fixed for other sessions and concepts . It is easy to use; no need to high level of training before use it by either the student or the teacher. Portability is one of its important characteristics because it could be created by standard software, SQL, PHP, etc.

The experiment we applied showed that the model gives high encouragement to the students to study by chat, he/she will be able to be comfortable with his peers rather discussing with his teacher with. At the same time the student can reflect his feeling about the discussion with his peer .This strategy of learning will motivate the student to study hardly and attend all the required chat sessions, because he will be evaluated and modeled for the next sessions.

The student will feel more relax when finishing the study concept by concept, and having more than chance to study the concept, one through the virtual classroom and one through discussion.

5.4.2 Limitations

Complex to design and implement, saying that we use standard software doesn't mean that the system can be constructed in short time . To implement these systems we need high level of experience in programming, network administration, database administration, web technology, education theories, and so on. Saying we need to construct adaptive e-learning chat it means we need a hard work to build the chat room and integrate it to the virtual classroom. Same problem emerges when we create the peer evaluation .We need experts in technology, education and psychology to have high level of questions which could get the right opinion of the student about his peer.

The students sometimes believe only in traditional class and real teacher, some of them but not the all do not feel confident when they study using chat with other students. The student may fail to study and have knowledge; this may cause him to discuss wrongly which will tag him with red colors in the next sessions and give bad impression by his peers about him.

The system did not introduce a way to deal with practical study and solving problems, and how to discuss these topics in the chat tool.

The system cannot have control about the student behavior during the peer evaluation process, let us say that the student may not be serious and answer the questionnaire however and this will give bad outcome from the system.

5.5 Conclusion and future work

The proposed model helps in virtual classrooms to expand the circle of its functions by opening the way for concepts discussion among students and having results on this subject. As we previously discussed, the discussion among students makes reliance on students and not on the teacher. To raise the level of the students by placing responsibility on their shoulders in the delivery of information to each others. But the problem was concentrated in how to assess the performance of the students during the session, which is reflected in their scores stored in their files. This process is done through various ways, namely:

1- Time model time: This model should work on the analysis of the use and exploitation of the session time in effective and useful way by the student.

2 Text model: This model should work on the analysis of student usage of his words during the session and the proximity or distance from the subject of the discussion .And the degree of usefulness for his peers.

3 Peer evaluation: This model gives the student the opportunity to evaluate his peers during the session; a right is used in many cases in educational process.

From the last three models we got an equation to evaluate the student during the chat session as discussed in chapter 4 . This equation could be used for any educational chat room.

The student modeling within the chat session is not only getting a score for the student but also having the student characteristics to add value for the adaptivity process which represents the core of the main study .Knowing the student characteristics will help the students to know more about his abilities and get better benefits from him. For this reason a new value is added to the chat room interface which describes the student abilities in time, text usage in the last session .His peer will rely on these characteristics to conduct the session with him .

It is clear that the field of adaptive chat room in e-learning is a highly complex and somewhat controversial area of research and one that has no quick solution. There does not seem to be any particular evidence to invalidate this area of research and any work carried out by others should not be dismissed out of hand.

The model should be enhanced to have better functions and results. We are going to perform the following functions in the future to have better performance:

1- Using fuzzy logic to get more precise results about the student activities.

2- Involving the adaptive multimedia resources in the chat session.

3- Using smart agents in the session to give instructions to the student according to the peer behavior during the session.

.

'

References ---

[1] Bush, V. As We May Think. The Atlantiv Monthly.

[2] A. Haake, "CoVer: a Contextual Version Server for Hypertext Applications", ECHT'92 Proceedings, ACM, 1992.

[3] Douglas Tudhope, Daniel Cunliffe, Joseph M. Reagle Jr.w, "New review of Hypermedia and multimedia", University of Glamorgan, UK, 2011.

[4] Advaith Siddharthan, "Adaptive Hypermedia ", Department of computing Science, University of Aberdeen, Scotland, UK, 2012.

[5] Peter Brusilovsky , "Adaptive Hypermedia for Education and Training " , School of Information Sciences, University of Pittsburgh,Pittsburgh, PA 15260, USA , 2012

[6] Sasakura , Yamasaki , "A framework for adaptive e-learning systems in higher learning

with information visualization ", IEEE ,2007 .

[7] Dafne Gonzalez ,'TEACHING AND LEARNING THROUGH CHAT:A TAXONOMY OF EDUCATIONAL CHAT FOR EFL/ESL', Universidad Simon Bolivar Caracas, Venezuela iatefl , 2004 .

[8] Jabari, N.A.M., Hariadi, M. & Purnomo, M. H. 'Intelligent Adaptive Presentation and

E-testing System based on User Modeling and Course sequencing in Virtual Classroom'.

IJCA 50 (9): 0975-8887. 2012.

[9] Nabil Arman, " E-learning materials development: Applying software reuse principles and granularity levels in the small ", Palestine polytechnic university, 2009.

[10] Haichao Dong, Siu Cheung Hui, Yulan He, (2006) "Structural analysis of chat messages for topic detection", Online Information Review, Vol. 30 Iss: 5, pp.496 ' 516

[11] Tsarouchas, V. & Economides, A. A,"The use of chat by Greek students", International Journal of Advanced Media and Communication, Vol. 2, No. 2, pp. 209-220, Inderscience. ISSN (Online): 1741-8003, ISSN (Print): 1462-4613. 9SJR= 0.038 (2007); H-index= 1] , 2008

[12] James M. Hudson, "THE ROLE OF CHATROOMS IN FACILITATING LEARNING BEHAVIORS IN SMALL GROUP DISCUSSIONS", PHD, Georgia Institute of Technology, 2006.

[13] Linan Zheng,"Instant Messaging: Architectures and Concepts Student Mentor", Institute of Communication Networks and Computer Engineering, University of Stuttgart, 2005.

[14] [Moodle , 2012] www.Moodle.com

[15] Kuan C. Chen ,' Text Mining e-Complaints Data From e-Auction Store With Implications For Internet Marketing Research 'Journal of Business & Economics Research , USA , 2009

[16] Siu Cheung Hui , Yulan He ,Haichao Dong ,'Text Mining for Chat Message Analysis' ,NTU , 2007

[17] (Source- Mailvaganam Hari, Text Mining for Fraud Detection- Creating cost effective data mining solutions for fraud analysis', 2007, http://www.dwreview.com/Data_mining/Effective_Text_Mining.html)

[18] Avan Wadia , Diana Parkinson ,'Consulting How to design and use free online surveys to collect feedback on your services' . Monitoring and evaluation consulting , 2010.

[19] Vivek Bhaskaran Co-Founder , How to effectively conduct an online survey, Question Pro , 2010.

[20] Dunwei Wen, Sabine Graf, Chung Hsien Lan, Terry Anderson, Kinshuk, and Ken Dickson , "Adaptive Assessment in Web-Based Learning.", ICME, page 1846-1849. IEEE, (2007)

[21] http://www.panalysis.com/resources/articles/understanding-time-on-site.aspx

[22] Istvan K. Nagy and Csaba Gaspar-Papanek,'User Behaviour Analysis Based on Time Spent on Web Pages'Department of Telecommunications and Media Informatics,Budapest University of Technology and Economics,2nd Magyar Tudosok Krt., Budapest, H-1117,2007

[23] "Time Spent on Site, Not Page Views Will Determine Web Rankings", http://www.itbusinessedge.com/cm/community/news/vam/blog/time-spent-on-site-not-page-views-will-determine-web-rankings/?cs=17757

[24] David A. Freedman, Statistical Models: Theory and Practice, Cambridge University Press (2005)

[25] Armstrong, J. Scott (2012). "Illusions in Regression Analysis". International Journal of Forecasting (forthcoming) 28 (3): 689. doi:10.1016/j.ijforecast.2012.02.001.

Allin Cottrell ,'Regression Analysis: Basic Concepts ' ,2007 [26]

[27] R. Dennis Cook; Sanford Weisberg Criticism and Influence Analysis in Regression,Sociological Methodology, Vol. 13. (1982), pp. 313'361

[28] Crina Grosan, Ajith Abraham , Rule -Based Expert Systems, Intelligent Systems

Intelligent Systems Reference Library Volume 17, 2011, pp 149-185 ,

[29] http://www.j-paine.org/students/lectures/lect3/node5.html

[30] Michael Negnevitsky ,'Arti'cial Intelligence ,A Guide to Intelligent Systems ,Second

Edition ,adison wisely , 2005 , p37

[31] Ajith Abraham ,'Rule-based Expert Systems ',Oklahoma State University, Stillwater, OK, USA ,2005

[N1] Jabari, N.A.M., Hariadi, M. & Purnomo, M. H ,'Agent Based Adaptive Educational ChatTool Model'. JATIT, 2013 .

[N2] De Bra, P., Ad Aerts, Berden, B., De Lange, B.,Rousseau, B., Santic, T., Smits, D., Stash,N.,"AHA! The Adaptive HypermediaArchitecture" Department of ComputerScience Eindhoven University of Technology,2003.

[N3] Liu , Tang , Ma ," A rule'based Approach forstudent modeling " , IEEE , 2008 , p 526-530 .

[N4] Boban Vesin , 'Rule-Based reasoning for building learner model in programming tutoring system '10th international conference on Advances in Web-Based Learning , Pages 154-163 ,2011

[N5] Dafne Gonzalez,'Teaching English with Technology, A TAXONOMY OF EDUCATIONAL CHAT FOR EFL/ESL ' , Teaching English with Technology vol. 3, no. 4, pp. 57-69, 2004.

[N6]ALFRED KOBSA , 'Generic User Modeling Systems ',User Modeling and User-Adapted Interaction 11: 49-63, 2001.

[N7] Msn.con, 2006.

[N8] Yahoo.com, 2006.

[N9]ICQ.com, 2006.

[N10] Davey ,Envall ,'Instant Messaging: Functions of a New Communicative Tool ', Anthropology 427 Doing Things With Words , Master thesis , 2004.

[N11] Jiawei Han and Micheline Kamber ,'Data Mining:Concepts and Techniques ',book,Second Edition,Elsevir ,University of Illinois at Urbana-Champaign,2006 .

[N12] Michelle Schwartz,'peer and Self Evaluation of Teaching',The learning and teachning office ,2008 .

[N13] Joel R. Evans and Anil Mathur,'The value of online surveys',Zarb School of Business, Hofstra University, Hempstead, New York, USA ,2005.

[N14] ABBAN, KENNETH,' UNDERSTANDING THE IMPORTANCE OF TIME MANAGEMENT TO ASSISTANT REGISTRAR'S IN THE REGISTRARS DEPARTMENT OF THE UNIVERSITY OF EDUCATION 'International Journal of Scientific & Engineering Research Volume 3, Issue 12, December-2012 .

[N15] Robert L. Mason , Richard F. Gunst ,' Statistical Design and Analysis of Experiments With Applications to Engineering and Science ', book, Second Edition, A JOHN WILEY & SONS PUBLICATION, Southern Methodist University ,2003.

[N16] Emmanuel G. BLANCHARD, Boris VOLFSON, Yuan-Jin HONG, Susanne P. LAJOIE ,'Affective Artificial Intelligence in Education: From Detection to Adaptation ',ATLAS Laboratory, McGill Faculty of Education, Montr??al (QC), CANADA ,2010.

[N17] Hyacinth S. Nwana ,'Intelligent Tutoring Systems: an overview ', Department of Computer Science, University of Liverpool,1990.

[N18] Peter Brusilovsky1, and Eva Mill??n2,'User Models for Adaptive Hypermedia and Adaptive Educational Systems ',School of Information Sciences ,University of Pittsburgh, University of Malaga ,2006 .

[N19] Martins, A. C., Faria, L., Vaz de Carvalho, C., & Carrapatoso, E. (2008). User Modeling in Adaptive Hypermedia Educational Systems. Educational Technology & Society, 11 (1), 194-207.

Source: Essay UK - http://doghouse.net/free-essays/information-technology/adaptive-hypermedia.php


Not what you're looking for?

Search:


About this resource

This Information Technology essay was submitted to us by a student in order to help you with your studies.


Rating:

Rating  
No ratings yet!


Word count:

This page has approximately words.


Share:


Cite:

If you use part of this page in your own work, you need to provide a citation, as follows:

Essay UK, Adaptive Hypermedia. Available from: <http://doghouse.net/free-essays/information-technology/adaptive-hypermedia.php> [22-02-19].


More information:

If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal:


Essay and dissertation help

badges