关键词: 查询似然模型, 链接模型, DocRank, 文本矩阵

Abstract: This paper analyzes query likelihood model, and the inherent relation between different texts is not considered in this model. To tackle the drawback, link model is introduced to the text retrieval and a new algorithm is proposed for calculating the?correlation?between texts（called DocRank algorithm）. By calculating?the correlation?between each two texts to form the matrix?of?texts and using?power iteration?method?to calculate the?priority value of each text. It is integrated into the query likelihood model in order to accurately locate the text which is supposed to find. The experimental results show that the proposed algorithm can be used effectively in text retrieval.

Key words: query likelihood model, link model, DocRank, texts matrix