INTELLIGENT AGENTS IN DISTRIBUTED MONITORING SYSTEMS Bogdan Patrut LAP Lambert Academic Publishing, ISBN 978-3-8383-9735-1 August 8, 2010 Click here to buy this book from Amazon.com

TABLE OF CONTENTS

Chapter 1 5

INTRODUCTION 5

1.1. Research Context 5

1.2. Book Objective 7

1.3. Personal contributions 8

1.4. Structure and organization of the book 11

1.5. Acknowledgment 12

Chapter 2 13

CURRENT STATE OF THE RESEARCH IN THE FIELD OF INTELLIGENT AGENTS 13

2.1. Agents, software agents and intelligent agents 15

2.2. Intelligent Agents Architectures 19

2.3. Artificial feelings of agents 24

2.4. Conversational and Pedagogical Agents 29

2.4.1. Conversational Agents 29

2.4.2. Pedagogical Agents 29

2.4.3. Projecting the interaction of character-like pedagogical and conversational agents 32

2.4.4. Advantages and disadvantages of using pedagogical agents 33

2.5. Ms Agent Technology 34

2.6. Agent-oriented Programming 37

2.7. From agents to multi-agent systems 39

2.8. Classification of multi-agent systems 40

2.9. Organizing the multi-agent systems 42

2.10. Formalisms for multi-agent systems 44

2.10.1. Logical formalisms for multi-agent systems 45

2.10.2. The problems of formalisms for multi-agent systems 48

2.11. Social abilities of intelligent agents within the multi-agent systems 50

2.11.1. Inter-agent communication languages 51

2.11.2. Negotiation and trust among agents 52

2.12. Internet Agents 53

2.13. Mobile Agents 54

2.14. Frameworks for multi-agents systems 55

2.15. Applications of the multi-agent system 57

2.15.1 Industrial applications 57

2.15.2. Other applications of the multi-agent systems 58

2.16. Internet – open and extensible distributed system 60

2.17. Limits to the multi-agent technology 64

2.18. Conclusions and personal opinions 65

Chapter 3 67

S-AGENTS AND MONITORING MULTI-AGENT SYSTEMS 67

3.1. Definitions 67

3.2. Graphic representation 71

3.3. Reducing normal s-agents 74

3.4. Examples of monitoring multi-agent systems 77

3.4.1. A development system for a hyper-encyclopedia 77

3.4.2. An educational software system 78

Chapter 4 80

MAGELAN – A MULTI-AGENT SYSTEM FOR MONITORING A HYPER-ENCYCLOPEDIA 80

4.1. The problem of building up an intelligent hyper encyclopedia 80

4.1.1. The intelligent hyper encyclopedia 80

4.1.2. Example of hyper-encyclopedia 82

4.1.3. The artificial trainer or the s-agent 85

4.2. The architecture of the MAgeLan system 87

4.2.1. The system of the s-agents 87

4.2.2. Structure and functioning of an s-agent 88

4.2.3 Intercommunication of agents and their contact with the software environment 90

Chapter 5 91

MONITORING THE EDITING ACTIVITY USING INTELLIGENT AGENTS THAT ARE SPECIALIZED IN NATURAL LANGUAGE PROCESSING 91

5.1. The WordNet Agent (WNA) 92

5.1.1. WordNet – an ontological net of meanings 92

5.1.2. Relationships between meanings 93

5.1.3. A library for accessing WordNet 94

5.1.4. Functioning of the WordNet Agent 96

5.2. The DEX (DXA) Agent 99

5.2.1 The Dex Online database 100

5.2.2. The way the Dex Online Agent functions 101

5.2.3. Using phonological alternations and inflectional types in the morphological analysis 102

5.2.3.1. The issue of the morphological analysis 102

5.2.3.3. The inflection tables 109

5.2.3.4. The algorithm of the DXA agent 111

5.3. The Wikipedia agent 111

5.3.1. References on Wikipedia 112

5.3.2. Functioning of the Wkipedia Agent 112

5.4. The Google Occurrences Agent 114

5.4.1. Google and PageRank 114

5.4.2. Generating words that have close web-contexts 115

5.5. The Word History Agent 120

Chapter 6 122

MONITORING THE WEB SEARCH BY USING INTELLIGENT AGENTS 122

6.1. The Semantic Rules Agent 122

6.1.1. Search rules 122

6.1.2. Bivalent verbs and problems of pragmatics 127

6.1.2.1. Bivalent verbs 127

6.1.2.2. The reasoning problem in social contexts 128

6.1.2.3. Some features of the bivalent verbs 130

6.1.2.4. A method for reasoning and obtaining the equivalence classes for the bivalent verbs 134

6.1.2.5. The values of signs, powers and implications 137

6.2. The Academic agent 138

6.3. The Hyperlinks Agent 142

6.3.1 The tasks of the LKA Agent 142

6.3.2. The HtmlDocument class 142

6.3.3. Functioning of the LKA Agent 144

6.4. The Internet History Agent 146

6.4.1. The tasks of the IHA agent 146

6.4.2. Structures of necessary data 146

6.5. Collecting, sending and assembling information 148

6.5.1. The Transfer & Execute Agent 148

6.5.2. The coordination agent and the conversational agent 152

6.5.3. An architecture of intelligent agent with temperament to CVA 154

6.6. Using the MAgeLan system and the results we got 159

6.7. Conclusions 162

Chapter 7 163

CONTEST – MULTI-AGENT SYSTEM OF MONITORING THE TRAINING 163

7.1. The architecture of the ContTest multi-agent system 163

7.2. The internal architecture of an s-agent 163

7.3. The initial and permanent assessment agent 165

7.4. The teaching agent and the exemplifying agent 166

7.5. The agent responsible for generating tests 167

7.6. The Checking Answers Agent 170

7.6.1. The relationship of confusion 171

7.6.2. Solving errors of accounting analysis 173

7.7 The final evaluation and grading agent 175

7.8. The coordination agent and the way an s-agent operates 175

7.9. Experimental data using the ContTest system 176

Chapter 8 181

CONCLUSIONS AND FURTHER RESEARCH GUIDELINES 181

REFERENCES 184

LIST OF FIGURES

Figure 2.1. The software agent 16

Figure 2.2. Classification of agents according to N. Keil 18

Figure 2.3. H. Nwana’s classification 18

Figure 2.4 The beliefs-desires-intentions architecture 22

Figure 2.5. Agents with vertical layers 24

Figure 2.6. Hierarchical organization 43

Figure 2.7. Organizing the agents in communities of experts 43

Figure 2.8. Group organization 44

Figure 2.9. Communicating a message 62

Figure 3.1. A simple s-agent having a blockage in A1 and an infinite cycle in A2 73

Figure 3.2. A normal s-agent and its reduced s-agent 74

Figure 4.1. An intelligent hyper-encyclopedia 82

Figure 4.2. A private encyclopedia 83

Figure 4.3. A three-student example. Articles already introduced and lists of tasks 84

Figure 4.4. Students and their trainers 85

Figure 4.5. The system of agents that form an artificial trainer (s-agent) for a student 86

Figure 4.6. The architecture of the MAgeLan system 88

Figure 5.1. COM objects and interfaces 95

Figure 5.2. Example where the WordNet agent is used 97

Figure 5.3. The structures of the two tables from the Dex Online database 100

Figure 5.4. The DXA agent used to get definitions, synonyms and antonyms of a word while editing in Word 101

Figure 5.5. The endings tree in reverse order, for masculine nouns 108

Figure 6.1. Three classes, two of them have a common element 126

Figure 6.2. Illustration of the symmetry of a bivalent interpersonal verb (to take a walk with) 131

Figure 6.3. Illustration of the transitivity of bivalent interpersonal verb (to be in the same place as) 131

Figure 6.4. Values of signs from the standpoint of agents and receivers 132

Figure 6.5. Representations of a class of bivalent verbs 133

Figure 6.6. Implication relationships within a class and towards a verb from another class 133

Figure 6.7. An example of HTML DOM (Document Object Model) 143

Figure 6.8. Graph structure associated with hyper-encyclopedia articles 149

Figure 6.9. The timer model – Ms Agent for a conversational agent 153

Figure 6.10. The textbox model – Ms Agent for a conversational agent 153

Figure 6.11. Expressing figures with the CVA conversational agent: (a) - discontentment, (b) - hope, (c) - contentment 158

Figure 6.12. Using the MAgeLan system on a network computer 160

Figure 6.13. The number of uses of the MAgeLan system for each of the four users 161

Figure 6.14. The graphic of using the MAgeLan system in the four fields 161

Figure 7.1. The architecture of a ContTest s-agent 164

Figure 7.2. The structure of the “applying the algorithm of the accounting analysis” s-agent 165

Figure 7.3. The initial and permanent evaluation agent – exemplification for the accounting analysis 165

Figure 7.4. Examples of “teachers” in video sequences 167

Figure 7.5. The test generating agent 169

Figure 7.6. A typical accounting problem where sums are randomly generated 169

Figure 7.7. The architecture of the Checking answers agent 170

Figure 7.8. Functioning of an s-agent 176

Figure 7.9. Checking an answer 178

Figure 7.10. Grading agent with human face expressing its discontentment towards the results the student reached 179

Figure 7.11. Grades obtained by subjects after applying the traditional and the ContTest training 180

LIST OF TABLES

Table 1.1. Book structure 11

Table 2.1. Merging of plans and the emotions through which they are expressed (according to Oatley92) 28

Table 2.2 Methods and actions for an Ms Agent character 36

Table 5.1. the summary of the search for the word “cepelor” 109

Table 5.2. The declination of the feminine nouns having different types of inflections (excerpt) 110

Table 6.2. An example of representation of notions about an agent with temperament 155

Table 6.3. The representations of the actions of the agent with temperament 156

Table 7.1. The matrix of the relationship of confusion where only 0 (no confusion) and 1(the confusion is made) values occur. (excerpt) 172

Table 7.2. The matrix of the relationship of confusion, where integral values appear, starting with 0 (excerpt) 173

Research Context

Intelligent agents and multi-agent systems are a modern technology with different applications (air traffic control, medicine, entertainment industry, military field, education). While the economic and social importance of the Internet increases, it is easily understood that most of the computer applications will soon become distributed systems.

Internet, especially the World Wide Web, is a great information resource, offering possibilities to get documented, informed, advertised, entertained, to communicate, all this permanently transforming the way in which people get information, study and do research, buy and do business, communicate and have fun.

Many of the current scientific research results, regardless of the approached field, are published on the web. On the web we find courses and laboratory work, technical manuals and documentation, free encyclopedias, conference papers and specialists’ or amateurs’ opinions on discussion forums. The web is also full of online linguistic resources and explanatory (monolingual) dictionaries, data collections on specific themes, bilingual dictionaries and many others.

The information and documentation resources found on the web are very well structured and are provided either by experts or teams of volunteers supervised by autocratic teams of experts. These resources have proven quality and can compete with most similar works, organized and edited by recognized academic or professional institutions. These online resources are daily accessed by many people directly from their computers and the results they get are often encountered in different documents by different authors. We make reference, for instance, to the Wikipedia free encyclopedia, the WordNet semantic database or the Explanatory Dictionary of the Romanian Language (monolingual dictionary), the online version, i.e. DEX Online. If WordNet is build up by a group of linguistic engineers, Dex Online and Wikipedia are achieved in a collaborative manner. The information quality from the latter two resources can often be questioned. However, it is known that there are groups of experts who regularly improve the quality of Wikipedia articles. A study of the Nature journal [Type06] shows that, despite the fact that Wikipedia is developed by volunteers, while Britannica Encyclopedia by experts, the number of errors encountered on a certain subject (biology, for instance) is about the same (116 vs. 123). Also, in [Blumen08], a model for determining quality articles is suggested.

Still on the web there is Google search engine, which is daily used by millions of people. We cannot neglect it. Google has become a sort of indicator for the visibility of something or someone in the virtual world of the Internet. Google is at the top of the search engine hierarchy. It can be used as an indicator of the “importance” of a web page related to certain terms that have been searched. The higher a web address gets placed in a Google search, the more important or relevant it is considered for a certain theme. However, Google also adopts a promoting policy on a commercial basis, but the right side is usually used to do so.

There are also other search engines, such as Yahoo, which is ranked the 2nd probably; many people use its free e-mail services.

Google Scholar, Scirus, SSRN, CiteSeer and others are thematic search engines. Google Scholar can find courses, articles and books for pupils and students. Scirus is mostly used in the academic field. SSRN (Social Sciences Research Network) is a network of search engines for the main fields of social sciences. One can find here articles and other materials from the fields of economics, psychology, sociology, law, political science and others.

Similarly, CiteSeer is a search engine specialized on computer science articles and related fields. The visibility of an article and its indexing on CiteSeer can point out its scientific importance at a given moment.

On one hand we have the web and all the information and linguistic resources available. It is explored by using navigation programs, such as Microsoft Internet Explorer or Mozzila Firefox.

On the other hand, we have nowadays a mass of people more or less engaged in research and drafting papers. Most of the time, they use a text editor such as Microsoft Word or OpenOffice Word.

In education (and not only), a large part of the assessment is based on thesis writing. Nowadays, if not everyone writes, at least all those who write, they do it directly on the computer. Drafting or editing a research paper (a thesis, a master thesis, a report or a PhD thesis) is done directly on the computer, often by using Microsoft Word.

The web is a distributed system. Wikipedia and Google Docs are examples of how team work, distributed work and distributed research can be effectively used.

Nowadays, almost everybody writes on the computer, in Word. Nowadays, almost everybody do Google searches or use Wikipedia to get documented.

On the other hand, today many people use computer-assisted training. However, we know that an educational software cannot prove intelligence or feelings. Therefore it is in the interest of many for us to discuss the computer systems that are able to prove such human abilities and integrate them in the educations software.

Although they seem to be completely different issues, assisting the editing activities and web searches, as well as the interactive training can be brought to a common denominator: monitoring the distributed systems. Thus, in both cases we deal with the issue of monitoring some distributed systems. We will create a general theoretical and practical tool to develop intelligent agents for such issues. In implementing agents, there will be several problems that will be solved using original techniques from the field of natural language processing, of emotional agents or using original algorithms on certain structures that can be modeled by graphs.

In the context of our research, the following questions arise:

· Can documentation services be automatized within online electronic resources?

· Can searches using search engines be automatized?

· Can online linguistic resources be automatically used?

· To what extent can distributed work, distributed team research be automatized?

or

· How can we organize agents within a “virtual teacher”?

· How can we endow a “virtual teacher” with feelings or temperament?

· How can we extract pragmatic information from different texts?

· How can we more effectively represent certain notions from a given field?

· Our book will try to answer such questions by creating a theoretical tool and implementing it as architectures of multi-agent systems.

Book Objective

The objective of our book is to create a general formal tool to describe agents that monitor distributed systems, as well as to develop two such systems. A
first system will be designed and developed for:

· monitoring and improving the effort of the researcher in his web documentation;

· monitoring and improving the effort of the researcher in drafting a paper;

· checking the originality of a scientific study;

· assisting a team of researchers in creating a smart encyclopedia.

The system will operate in an Intranet network, connected to the Internet. The multi-agent system will be organized in groups of agents, to each group being associated with a Network node and a user. Each group will be made out of several agents, each of them having to complete specific tasks. There will be a coordinating agent that will work directly with the user, communicating in a natural way; it will also have to provide the necessary resources to the other agents. Moreover, a transmission agent will be communicating with other nodes so as to help one another and achieve the common goal.

A second system will be designed and developed to become a powerful and interactive training tool. The system will be organized in groups of agents, each one having to teach a certain lesson. Each agent from each group will have special educational tasks. By using the multi-agent system it will be possible to train students in accounting.