I.

INTRODUCTION: With the rapid growth of Internet and World Wide Web (WWW), we have now entered into a new information age. The Web provides a total new media for communication, which goes far beyond the traditional communication media, such as radio, telephone and television. The Web has significant impacts on both academic research and ordinary daily life. It revolutionizes the way in which in-formation is gathered, stored, processed, presented, shared, and used. The Web offers new opportunities and challenges for many areas, such as business, commerce, marketing, finance, publishing, education, research and development. For computer scientists, the Web introduces many new research topics and provides a new platform to reconsider old problems. It might be high time to create a new sub-discipline of computer science covering theories and technologies related to the Web.The term Web Intelligence was born in a paper written by Ning Zhong, Jiming Liu Yao and Y.Y.Ohsuga in the Computer Software and Applications Conference in 2000. In the fore going section of this chapter following information is provided : definition of Web Intelligence in section 2. Characteristics of intelligent web in section 3. In section 4, the four levels of Web intelligence is presented. The next section is about the social intelligence need.In section 6 the research agenda of WI is presented. In Section 7, necessity and beneficiary to have a new sub-discipline of computer science labelled by WI is illustrated. In Section 8, an overview of Artificial Intelligence and its relevance to WI is shown. II.WHAT IS THE WEB INTELLIGENCE? According to Ning Zhong, Jiming Liu Yao and Y.Y.Ohsuga [1]: Web Intelligence (WI) exploits Artificial Intelligence (AI) and advanced Information Technology (IT) on the Web and Internet. This definition has the following implications. The basis of WI is AI and IT. The I happens to be shared by both AI and IT, although with different meanings in them, and W defines the platform on which WI research is carried out. The goal of WI is the joint goals of AI and IT on the new platform of the Web. That is, WI applies AI and IT for the design and implementation of Intelligent Web Information Systems (IWIS). Integration of IT with AI Applications of AI on the Web

Figure 1: What is Web Intelligence? An IWIS should be able to perform functions normally associated with human intelligence, such as reasoning, learning, and self improvement. There perhaps might not be a standard and noncontroversial definition of WI, as the case that there is no standard definition of AI. The term, Web Intelligence, should be considered as an umbrella or a label of a new branch of research

centered on the Web. The definition simply states the scopes and goals of WI. This allows us to include any theories and technologies that either fall in the scopes or aim at the same goals. WI will be an ever-changing research branch. It will be evolving with development of the Web as new media for information gathering, storage, processing, delivery and utilization. Although no one can predict the future in detail and without uncertainty, it is clear that WI would have huge impacts on the application of computers, which in turn will affect our everyday lives. III. THE INTELLIGENT WEB Despite current technological advances, it is hard to predict what the Webs next paradigm shift will be. However, this change will transform the Web into an intelligent entityhence, the term Web intelligence. The next-generation Web will go beyond improved information search and knowledge queries and will help people achieve better ways of living, working, playing, and learning. To fulfill its potential, the intelligent Webs design and development must incorporate and integrate several fundamental capabilities.[2] 3.1 Autonomic Web support The intelligent Web functions essentially as an autonomic entity. The Web automatically regulates the functions and cooperation of related Web sites and available application services. Making the Web autonomic requires several challenges to meet: Reflexive server propagation. An intelligent Web server must be able to automatically self delegate its functional roles to other services, along with its corresponding spatial or temporal constraints and operational settings. Specialization. An intelligent Web server must itself be an agent specialized in performing some roles in a certain service. The association of its roles with any service will be measured and updated dynamically. Growth. The intelligent Web agent population will change dynamically as agents selfrepro-duce to become more specialized or as agents deactivate. Autocatalysis. Catalysis is a reaction between two or more entities precipitated by a separate agent. As search requests activate intelligent Web agents, they will evolve through specialization to fill various roles, generating associations with some services and among themselves that will aggregate auto catalytically. 3.2 Representation Intelligent Web agents can use the Problem Solver Markup Language (PSML) to specify their roles, settings, and relationships with any other services. The intelligent Web must also have the ability to process and understand natural language. It must understand and correctly judge the meaning of concepts expressed in words, such as good, best, and season. Further, the

intelligent Web must grasp the granularities of these terms corresponding subjects and the location of their ontology definitions. 3.3 Self-direction and learning In addition to the semantic knowledge that an intelligent search can extract and manipulate, intelligent Web agents must also incorporate a dynamically created source of metaknowledge that deals with the relationships between concepts and the spatial or temporal constraint knowledge that planning and executing services use. This allows the agents to self-resolve their conflicts. To solve specific problems, intelligent Web agents must be able to plan. The planning process uses goals and associated sub goals, as well as constraints.In the intelligent Web, ontologies alone will not be sufficient. 3.3 Personalization The intelligent Web can personalize interactions by remembering a particular users recent encounters and relating the topics and sites that a user accesses during different online sessions. It may further identify other goals and courses of action as a users interactions broaden and deepen, providing ever more data upon which to base its recommendations. As part of its personalized approach to user services, the intelligent Web will interact with the user when executing these tasks. In summary, semantics contributes a vital aspect to the intelligent Web. IT is expected that the Web to extend not only the knowledge of artificial assistants, but also their intelligence. IV.WIS FOUR LEVELS One can study Web intelligence on at least four conceptual levels, ranging from the lower, hardware-centered level to the higher, application-centered level. This framework builds upon the fast development and application of various Web technologies.[2] Internet-level communication, infrastructure,and security protocols. At its core, the Web is a computer-network system. WI techniques for this level include Web data prefetching systems built upon Web surfing patterns to resolve latency issues. The intelligence of the Webs prefetching routines comes from an adaptive learning process based on observations of user surfing behavior. Interface-level multimedia presentation standards.The Web functions as an interface for human-Internet interaction. At this level, the Web interfaces require adaptive crosslanguage processing, personalized-multimedia-representation, and multimodal-dataprocessing capabilities. Knowledge-level information processing and management tools. The Web serves as a distributed data and knowledge base. Accessing and manipulating this information requires semantic markup languages to represent the Webs contents in machineunderstandable formats. Agent-based autonomic computing functions such as searching,

aggregation, classification, filtering, managing, mining, and discovery can then use this data. Application-level ubiquitous computing and social intelligence environments. The Web can form the basis for establishing social networks that contain communities of people, organizations, or other social entities. Social relationshipssuch as friendship, co working, or exchanging information about common interestsconnect these entities. The study of WI thus encompasses issues central to social network intelligence. Users access the Webs multimedia content from stationary desktop computers and increasingly from mobile platforms as well.5 Ubiquitous Web access and computing from various wireless devices requires even greater adaptive personalization. WI should suit these needs well by providing techniques for use in constructing interest models derived from implicit inferences based on user behavior. V. SUPPORTING SOCIAL INTELLIGENCE NEEDS WI researchers must study both centralized and distributed information structures. The Webs information and knowledge can be organized in two ways: globally distributed within multilayers over the Webs protocol infrastructure, or locally centralized on an intelligent portal that provides Web services joined to its own cluster of specialized intelligent applications. Each approach, however, has a serious flaw: The global semantic Web approach faces combinatory complexity limitations, while the intelligent-portal approach limits uniformity and access. [3] One way to solve this problem is to use PSML for collecting globally distributed content and knowledge from Web-supported social networks. The collected data can then be incorporated with locally operational knowledge-databases to provide a local enterprise or community with centralized, adaptable, Web-intelligent services. Social networks create a self-organizing structure of users, information, and expert communities.[3] Such networks can play a crucial role in combining next-generation enterprise portals and Web search engines with functions such as data mining and knowledge management to discover, analyze, and manage social-network knowledge. The social network resides atop the four-level WI framework and benefits from support functions drawn from all levels of Web intelligence, including security, prefetching, adaptive cross-language processing, personalized multimedia representation, semantic searching, aggregation, classification, filtering, managing, mining, and discovery. VI. WI RESEARCH AGENDA WI research incorporates knowledge from existing disciplines, such as artificial intelligence and

information technology, in a totally new domain. At the same time, WI research also enriches these established disciplines as it introduces new topics and challenges. Figure 2 shows a schematic diagram of interrelated research topics from a Web-based, business-centric perspective. Researchers have achieved much in the field of Web intelligence, but several important issues provide the basis for the ongoing WI research agenda. [1] 6.1 Web mining With the rapid growth of information stored on the Web, researchers can use data mining to explore many different resources. Web mining applies data mining techniques to large Web data repositories such as documents, logs, and services. Web mining research thus straddles the crossroads of research undertaken by the database, information retrieval, machine learning, and natural-language processing communities, among others. 6.2 Automatic ontology construction Ontology is a key enabling technology for developing the semantic Web. It plays a crucial role in Web-based knowledge processing and management, such as Web community communication, semantic-based agent communication, knowledge based Web information retrieval, Web content understanding, and Web community discovery. Comprehensive ontology construction and learning, an active field of study for at least a decade, still requires more automatic methods. The tedious, cumbersome task of building an ontology manually remains a bottleneck that severely slows the development of semantics-based Web intelligence. 6.3 PSML and Web inference engine Distributed inference engines form PSMLs core. These engines can perform automatic reasoning on the Web by incorporating autonomically collected and transformed content and metaknowledge into locally operational knowledge and databases. A feasible way to implement PSML is to use an existing Prolog-like logic language supplemented with agents that perform dynamic-content updates, meta knowledge collection, and transformation tasks.

6.4 Social network intelligence The social intelligence approach to Web computing presents new opportunities for WI research and development. As the Web becomes an integral part of our society, WI can and should support Web-based social networks at all levels. Study in this area must receive as much attention as Web mining, Web agents, ontologies, and related topics. 6.5 Web-based computing The intelligent Web seeks to provide not only a medium for seamless information exchange and knowledge sharing, but also the sort of human-crafted resources that encourage sustainable knowledge creation and scientific and social evolution. The intelligent Web will rely on Grid-like service agencies that self-organize, learn, and evolve their courses of action to perform service tasks and transform their identities and interrelationships in communities. These services will also cooperate and compete among themselves to optimize their resources and utilities and those of others. 6.6 Benchmark applications To effectively develop and evaluate systems and applications that address WI research issues, one must consider benchmark applications that will demonstrate these capabilities. Suppose we want to conduct a Web-based search to compile the data and generate a market report for an existing product or a potential new product. To perform these tasks, an information agent will mine and integrate available Web information, which will in turn be passed to a market analysis agent. The analysis will involve the quantitative simulation of customer behavior in a marketplace, instantaneously handled by other service agencies involving a large number of Grid agents. Given that the number of variables can number in the hundreds or thousands, generating one prediction can easily require significant computer resources

Figure 3: Example of an advanced WIS. VII. MOTIVATIONS AND JUSTIFICATIONS FOR WI

The introduction of Web Intelligence (WI) can be motivated and justified from both academic and industrial perspectives. Two features of the Web make it a useful and unique platform for computer applications and research, the size and complexity. The Web contains a huge amount of interconnected Web documents known as Web pages. For example, the popular search engine Google claims that it can search 1,346,966,000 pages as of February 2001. The sheer size of the Web leads to difficulties in the storage, management, and efficient and effective retrieval of Web documents. The complexity of the Web, in terms of connectivity and diversity of Web documents, forces us to reconsider many existing information systems, as well as theories, methodologies and technologies underlying those systems. One has to deal with a heterogeneous collection of structured, unstructured, semi-structured, interrelated, and distributed Web documents consisting of texts, images and sounds, instead of homogeneous collection of structured and unrelated objects. The latter is the subject of study of many conventional information systems, such as databases, information retrieval, and multimedia systems. To accommodate the needs of the Web, one needs to study issues on the design and implementation of the Web-based information systems by combining and extending results from existing intelligent information systems. Existing theories and technologies need to be modified or enhanced to deal with complexity of the Web. Although individual Web-based information systems are constantly being deployed, advanced issues and techniques for developing and for benefiting from the Web remain to be systematically studied. The challenges brought by the Web to computer scientists may justify the creation of the new sub-discipline, WI, for carrying out Web-related research.The Web increases the availability and accessibility of information to a much larger community than any other computer applications. The introduction of Personal Computers (PCs) brought the computational power to ordinary people. It is the Web that delivers more effectively information to everyone at finger tips. The Web, no doubt, offers a new means for sharing and transmitting information unmatchable by other media. The revolution started by the Web is just beginning. New business opportunities, such as e-commerce, e-banking, and e-publication, will increase with the maturity of the Web. It can hardly over emphasize more impacts of the Web on the business and industrial world. The creation of a new sub-discipline devoted toWeb related research and applications might has a significant value in the future. [4] VIII. PERSPECTIVES OF WI As a new branch of research, Web Intelligence exploits Artificial Intelligence (AI) and Information Technology (IT) on the Web. On the one hand, it may be viewed as applying results from these existing disciplines to a totally new domain. On the other hand,WI may also introduce new problems and challenges to the established disciplines. WI may also be viewed as an enhancement or an extension of AI and IT. It remains to be seen if WI would become a sub-area of AI and IT or a child of a successful marriage of AI and IT. However, no matter what happens, studies on WI can benefit a great deal from the results, experience, success and lessons of AI and IT.In their very popular textbook, Russell and Norvig [5] examined different definitions of

artificial intelligence from eight other textbooks, in order to decide what is exactly AI. They observed that the definitions vary along the two dimensions. One dimension deals with the functionality and ability of an AI system, ranging from thought processes and reasoning ability of the systems to the behavior of the systems. The other dimension deals with the designing philosophy of AI systems, ranging from intimating human problem solving to making rational decision. The combination of the two dimensions results in four categories of AI systems adopted from Russell and Norvig [5]:

This classification provides a basis for the studies of various views and approaches for AI. It also clearly defines goals in the design of AI systems. According to Russell and Norvig [39], they correspond to four approaches, the cognitive modelingapproach (thinking humanly), the Turing test approach (acting humanly), the the laws of thought approach (thinking rationally), and the rational agent approach (acting rationally). The two rows for separating AI systems in terms of thinking and acting may not be a most suitable classification. Action is normally the final result of a thinking process. One may argue that the class of systems acting humanly is a super set of the class of system thinking humanly. In contrast, the separation of human-centered approach and rationality-centered approach may have a significant implications in the studies of AI. While earlier research on AI was focus more on human-centered approach, rationalitycentered approach received more attention recently [5] The first column is centered around humans and leads to the treatment of AI as an empirical science involving hypothesis and experimental confirmation. A human-centered approach represents the descriptive view of AI. Under this view, a system is designed by intimating the human problem solving. This implies that a system should have the usual human capabilities such as knowledge representation, natural language processing, reasoning, planning and learning. The performance of an AI system is measured or evaluated through the Turing test. An system is said to be intelligent if it provides human level performance. Such a descriptive view dominates the majority of earlier studies of expert systems, a special type of AI systems. The second column represents the prescriptive or normative view of AI. It deals with theoretical principles and laws that an AI system must follow, instead of intimating humans. That is, a rationalist approach deals with an ideal concept of intelligence, which may be independent of human problem solving. An AI system is rational if it does the right thing and makes the right decision. The normative view of AI based on the well established disciplines such as mathematics, logic, and engineering. The descriptive and normative views also reflect the experimental and theoretical aspects of AI research. The experimental study represents the descriptive view. It covers theories and models for the explanation of the workings of the human mind, and applications of AI to solving problems that normally require human intelligence. The theoretic study aims at the development of theories of rationality, and focuses on the foundations of AI. The two views are complementary to each other. Studies in one direction may provide valuable insights into the other.

Web Intelligence concerns the design and development of intelligent Web information systems. The previous framework for the study of AI can be immediately applied to that of Web Intelligence. More specifically, we can cluster research in WI into the prescriptive approach and the normative approach, and cluster Web information systems in terms of thinking and acting. Various research topics can be identified and grouped accordingly. Like AI, a foundation of WI can be established by drawing results from the following many related disciplines: Mathematics: computation, logic, probability. Applied Mathematics and Statistics: algorithms, non-classical logics, decision theory, information theory, measurement theory, utility theory, theories of uncertainty, approximate reasoning. Psychology: cognitive psychology, cognitive science, human-machine interaction, user interface. Linguistics: computational linguistics, natural language processing, machine translation. Information Technology: information science, databases, information retrieval systems, knowledge discovery and data mining, expert systems, knowledge-based systems, decision support systems, intelligent information agents.The topics under each entry are only intended as examples. They do not form an exhausted list. In the development of AI, it was witnessed that the formulation of many of its new sub-branches, such as knowledge-based systems, artificial neural networks, genetic algorithms, and intelligent agents. Recently, non-classical AI topics have received much attention under the name of computational intelligence. Computational intelligence focuses on the computational aspect of intelligent systems . The application of AI in other disciplines also leads to new techniques in the corresponding fields. For instance, Business Intelligence (BI) is a result of applying artificial intelligence to the business domain. Artificial Intelligence in Medicine also proved to be a successful application. When viewing WI in such settings, one can identify at least two of its roles. WI may be interpreted Web based Artificial Intelligence as the study of particular aspects of AI in the context of the Web, in parallel to the study of computational intelligence. WI may also be interpreted as Artificial Intelligence on the Web which regards it as a new application of AI. A more practical goal of WI is the design and implementation of intelligent Web information systems (IWIS). It should be realized that an IWIS is an integrated system containing many sub-systems. To design such a system, it is necessary to apply a variety of theories and technologies. In his work on vision, Marr, convincingly made the point that a full understanding of an intelligent system involves explanations at various levels. The same argument is applicable to the development of an IWIS. One can identify at least two levels, the conceptual formulation and physical implementation. The conceptual formulation deals with foundations of IWIS, while physical implementation concerns with construction of an IWIS. The former depends on mathematics and logic, and the latter depends on algorithms and programming. Each level may be further divided into more sub-levels. Research in WI should

include any topics at different levels. [4] IX. SUMMERY While it may be difficult to define what exactly Web Intelligence (WI) is, one can easily argue for the need and necessity of creating such a subfield of study in computer science. With the rapid growth of the Web, we foresee a fast growing interest in Web Intelligence. Roughly speaking, we define Web Intelligence as a field that exploits Artificial Intelligence (AI) and advanced Information Technology (IT) on the Web and Internet. It may be viewed as a marriage of artificial intelligence and information technology in the new setting of the Web. By examining the scope and historical development of artificial intelligence, Here some fundamental issues of Web Intelligence in a similar manner is discussed. There is no doubt in our mind that results from AI and IT will influence the development of WI. In the foregoing 2 chapters the research agenda for WI about Semantic Web and Social network intelligence is discussed in detail. At the end the trends and challenges of WI research is also presented in brief

References:
1. N. Zhong et al., Web Intelligence (WI), Proc. 24th Intl Computer Software and Applications Conf. (COMPSAC 2000), IEEE CS Press, Los Alamitos, Calif., 2000, pp. 469-470. 2. 1. J. Liu et al., The Wisdom Web: New Challenges for Web Intelligence (WI), published in J. Intelligent Information Sys, vol. 20, no. 1, 2003. 3. H.P. Alesso and C.F. Smith, The Intelligent Wireless Web, Addison-Wesley, Reading, Mass., 2002. 4. Y. Yao et al., Web Intelligence (WI): Research Challenges and Trends in the New Information Age, Web Intelligence: Research and Development, Lecture Notes in Articial Intelligence 2198 (LNAI 2198), N. Zhong et al., eds., Springer-Verlag, Heidelberg, 2001, pp. 1-17. 5. Russell, S. and Norvig, P. Artificial Intelligence, A Modern Approach, Prentice Hall, 1995.