Study On Personalized Agent Technology And Its Implementation

As the lntemet'intranet is widely used, more and more information is saved as electronized style while the means for accessing information can't catch up with the increase of the information, which leads to information disorientation and information congestion.. Although many tools have been developed, the traditional information service systems such as Yahoo can't understand precisely what the users really want. For example, some information that the user isn't interested in is always listed in the front when the user searches for artificial intelligence knowledge. These systems ignore the user's personalized information demands and can't provide long-term active information services. Therefore, it is urgent for researchers to research and develop the intelligent information service system that can fully understand what the user is really interested in.Information aQent technology can help to understand the user's long-term information requirements in an intelligent way. By combining of such technologies as information retrieval. automatic classification, machine learning and in formation push, it aims to provide users \\'ilh accurate, reliable and efficient information services. Nowadays people are extensively researching information retrieval, information filtering and website navigation.This paper involves development of "Tian Luo Personal Information Agent (TLPIA)'. an active information push software. Its technology basis, implementation and performance are also presented. TLPIA is the important part of Techlinerfechlore Intelligent Information Navigation Platform, which is one of the projects of National 863 Programme The author took part in the R&D of this project and was responsible for the realization of the Feedback Information Collector module which is the key part to implement personalization.The contribution of the paper lies in: a)a more practical and more flexible model of '~K\V data is presented; b)TLPLA not only designs a relevant feedback procedure with the filtering ability that is optional to users, but also provides automatic tracking and learning function. The s~'stem turns the information coming from feedback information or learning from tracking into personalized vectors which are applied to further filter by calculating the distance between the personalized vector and the document through the decision tree ,and therefore it is a better way to resolve the problem of information congestion.The result of the experiment shows increased space-time performance and high information service quality.The paper also analyses some drawbacks in this system. At the end of the paper, some research aspects and improved techniques that will be focused in near future are discussed.The s"stem is now under use in some units and departments.