The Research On The Information Retrieval Based On The Markov Network

Along with the quick development of the internet, the information increases quickly everyday. How to obtain the information that customers want from this enormous resource databases is an important problem for all users. The search engines are the most valid means at present. Traditional retrieval models only make a simple match between a query and words in the documents, their effects are not so good influenced by uncertain factors like the relevance between the words. The intellectualized information retrieval has become a major research topic because the traditional information retrieval can not meet the requirements of users.We analyses Markov network retrieval model ,it integrates computer,topology and possibility which is a powerful tool for representation and inference of possibility knowledge. We consider the retrieval model as an inference process. Its isotropy take the query as the evidence source, the relevance document is looked upon as active document, improving the performance of information retrieval by adding useful information into retrieval process.After listing the research background and the general theory of the Markov network in this article, we summing up about four information retrieval models based on the Markov network. By using learning mechanism, analyzing and studying of corpus we use the correlation between words and construct a Markov knowledge network. Making use of edges without the direction in the network, activating the words closely related with the query as the additional evidential sources into the model. Our experiments show that the IMNM is highly applicable, and get optimal results in the retrieval of English documents. Comparing with the Bayies network model and the BM25 model, our model makes some improvements in retrieval performances. We have carried out several the following aspect job mainly:1 Summing up about information retrieval models based on the Markov network;2 Applying the reinforce algorithm to compute the relationship between words to construct the Markov Network and using multilayer inference process;...