Improving Information Retrieval within Language Model Framework by Integrating Adjacent and Distant Relation

Previous research has shown that using term relationships within language model could improve the effectiveness of information retrieval. But all models only consider adjacent relation or distant relation, none of them combine the two types relation. So in this paper, we propose a new dependency language model for improving information retrieval. In our model, phrases and co-occurrence terms are integrated into language model which includes adjacent relation and distant relation in a document. We test the new model on five TREC test collections and compare the results with the baseline model. The results validate the hypothesis that integrating adjacent and distant relation into language model can improve the performance of information retrieval.