Abstract: An idea is presented about users' personalized information which can be stored by a word-correlativity network, and the way to enable this network to be set up automatically in user's searching process is proposed according to the user's preference. To re-rank the results of queries as response from underlying search engine, three strategies were developed using the word-correlativity to re-evaluate them. Then, a personalized search algorithm using word-correlativity model is presented. A prototype system is thus designed to test the relevant performance of the algorithm. The testing results show that the word-correlativity model can store and form the correlations between key words of a text and verify the correctness of the retrieval algorithm.