A Concept Base Model For Intelligent Searcher And Its Document Evaluation

Search engine has greatly improved the availability of the information on Internet. But the present SE has also its disadvantages in searching and organizing the data that users required and the resultant high price on data fetching and reusing, because it is built on the technology of textual word matching within the whole document without comprehending and cannot catch the users' real search intents completely and accurately.So is intelligentized SE proposed. After analyzing the status of the SEs present and some possible solutions, this paper suggests that intelligentizing is the most effective and fundamental way to reach SE's objects, including recall, precision, speed and economization, especially the object of precision. More deeply, it also gives the reader some detail techniques to build an intelligent searcher, and the idea to present and handle the knowledge involved in the searching based on concept, concept relations and concept base. As we know, a searcher is the core part of a search engine and does all the routines of selecting and organizing documents for its users.Such a concept base is very suitable to retrieval on Internet, in that it inherits all the merits of some classification standards and knowledge-process solutions such as HNC and HowNet, and takes some transformations specific to web-searching. A concept base has also a very perfect behavior in helping text content representing, document comprehending, feature extracting and document matching. With the help of user-specific concept hierarchy and its application strategies, the concept operations based on concept relations can give a great support to construct user request, to analyze the statement and get the real search intent that is not stated clearly, and to evaluate the degree a document satisfies the request. They also make it possible as a more personalized and intelligent human-machine interface.In the fuzzy model of concept, concept base and concept operation, a concept is made up of and represented in its attributes, behaviors and relations with other concepts. All the fuzzy relations connecting concepts give the tips of the knowledge hierarchy in a man's mind in a given environment. And every relation takes its own role and has specific rules in aiding to learn user's background, hypothesize real search intent, represent text content, analyze document feature, and search and evaluate document.In order to implement the system, some suggests are proposed in the later part, especially those on the interface of human-machine interactivities and coral document feature process. Some protype systems show that such a concept-based comprehending mechanism makes a searcher more efficient and adaptive.