16 IEEE POTENTIALS 0278-6648/05/$20.00 © 2005 IEEE Intelligent agents are software pro- grams that act on behalf of human users or other systems to carry out arduous information gathering and processing tasks. Thus, an agent-oriented approach is well suited for helping “a human” to find needed information on the Internet. Some important multiagent applica- tions on the Internet are e-mail systems, teleconferencing, webwander- ers, and others that are sup- ported by Internet-scale multi- agent systems. In multiagent systems, agents should be able to interact with each other and with a changing external environment in an adaptable manner. This ability is part perception (the receiving of messages) and part action (the sending of messages). In a purely computer-based agent, these may be the agent’s only per- ceptual and acting abilities. Existing search and retrieval engines provide limited assistance to users in locating relevant information: search engines do not learn from previous searches so they do not adjust themselves to their users. In other words, search engines rank retrieved documents from highest to lowest based on fixed criteria (e.g., include all of the search terms in the document, number of times the terms appear) not based on users’ actions. However, autonomous, intelligent agents would transform passive search and retrieval engines into active, personal assistants according to Jansen because agents can adjust themselves to the pref- erences and wishes of individual users, by learning from performed tasks. Agents are able to continuously monitor the interested URL for new updates. Thus, intelligent software agents could be a promising solution to the current (threat of an) information over- load on the Internet. Activities on the Internet can be split into three layers: the demand side , the supply side, and the intermediaries. On the demand side (of information), i.e., the information searcher or user, the agent’s tasks are to find out exactly what users are looking for, what they want, if they have any preferences with regard to the information needed. On the supply side (of information), i.e., the individual information sources and suppliers, an agent’s tasks are to make an exact inventory of the kinds of services and information that are being offered by its supplier and to keep track of newly added information. The intermedi- aries mediate between agents (of the other two layers), i.e., act as (infor- mation) intermediaries between (human or elec- tronic) users and suppliers. The functionality of an agent would be maximally utilized if it was employed in such a three-layer structure. Intelligent agents would be able to help users not only with the search and filtering of information but also with cate- gorization, prioritization, selective dissemi- nation, annotation, and (collaborative) sharing of information and documents. According to Hermans, the functionality of the agent is maximized in the middle layer. The middle layer (agent) functions are: dynamically matching user demand and provider’s supply in the best possible way, unifying and possibly processing suppliers’ responses to queries to pro- duce an appropriate result, and current awareness, i.e., actively notify users of information changes, and bring users and suppli- ers together. Another ap- proach is a system called the Info Agent proposed by D’Aloisi and Giannini. It supports users in retrieving data residing in distributed and heterogeneous archives and repositories. This agent sys- tem is like a personal assistant responsi- ble for handling the user’s needs and for connecting the user with whoever can help solve the problem. Since at least the 1990s, researchers have predicted that one day nearly all the major Internet sites would be capable of hosting and willing to host some form of mobile code or mobile agents. For exam- ple, General Magic with its Telescript was a pioneer in mobile code/mobile compu- tation technologies <http://www.genmag- ic.com\>. Two other mobile agents and mobile-objects-related work illustrate the rapid growth of the field: IBM Japan’s Aglet <http://www.trl.ibm.co.jp/aglets> and ObjectSpace’s Voyager <http://www. objectspace.com>. The middle-layer approach Our proposed system would be a mid- dle-layer agent system working as a go- between for the user and the search engines. Figure 1 shows the proposed sys- tem’s scope. It would provide a unified environment with several agent types. These intelligent agents would interoperate to help the user get the needed information as fast and easily as possible. Text document categorization First, we need a text doc- ument categorization tech- nique to extract the features from the documents of interest to the user. From these features we draw a user search strategy. In our approach, we model text documents as transactions where items are words (features) from the Mostafa M. Syiam, Hamdy M. Kelash, Mohammed A. Berbar, Marwa A. Radad, and Hossam M. Faheem ©2001 PHOTODISC to help with the research Getting agents