Abstract:The core work of information retrieval including document classification and ranking operations, how to effectively compute the term weight of every document is one of a key technology. Use of the word co-occurrence relationship to create a text graph for each document, based on the idea of the importance of interaction between adjacent words, combining the characteristics of the word document word frequency characteristics, we iteratively compute weighting of each word. Further combining the global properties of text graph, such as density, we could rank the results of information retrieval. Experiments confirmed that the algorithm in standard data sets with good results.