Associated With The Theme Of Personalized Retrieval Model

With the rapidly increasing in the amount of network data, data retrieved by search engine has been more and more, and then it becomes more and more difficult to find the right results for people. But also the results are the same for different person whose search term is the same, which requires a personalized information retrieval model to meet the Individual requirements. Present, personalized information retrieval research is still in the initial stage, and it basically uses the user model to filter the results and resets the results. A lot of personalized retrieval system using artificial intelligence to implement, and make use of its characteristics to update the user model.This paper proposes the combining method of subject-type documents and user interest model to implement retrieval system based on studying the key technologies of personalized retrieval: In the user interest model organizations, it uses combining structure of linear structure and tree structure to construct the user interest model , and uses a XML file store user interest model; In updating the user interest model, it brings in the TCP/IP congestion control mechanism, and defines the concept of long-term or short-term interests according to the law of oblivion, and shows a type of update function and update algorithm, at last the experiments'data proves the validity of the update mechanism; In text similarity, it thinks about the difference of the weight of word to calculate the similarity word similarity; In feature extraction of text, it adds up the document structure factor based on TFIDF arithmetic to calculate; In the text classification, at first it study the current text classification algorithm, and then uses the word similarity to expand Naive Bayes algorithm, at last the experiments'data proves the validity of this algorithm.On the basis of overall design, this paper shows the overall framework design of personalized information retrieval system and implement of the user interest templates, updating mechanism, feature extraction, text classification and filtration of the retrieval results. At last, it illustrates the effectiveness of the system through the performance experiment of the personalized information retrieval system and experiment about impact of user model on the retrieval results with of recall ratio and precision ratio.