387 Vol. 50 No. 1 总第 期 ， drift the subject and focus on old web pages which cannot meet the actual needs of users. To improve this phenomenon an im⁃ ， ， proved PageRank algorithm is proposed. The algorithm uses the BM25 similarity algorithm to calculate the topic similarity assign ， different impact weights based on the similarity score and can increase the ranking of highly similar web pages using the number ， ， of times a web page is searched in the search engine cycle to represent the Web length of time that exists compensating for the ， weight of new pages. Introducing them into the PageRank algorithm makes the calculation of page PR values more reasonable. Exper⁃ iments show that the improved PageRank algorithm can effectively reduce the topic drift in search applications and the search re⁃
