In this presentation, I will describe a novel dependency language modeling approach for information retrieval. The approach extends the existing language modeling approach by relaxing the independence assumption. Our goal is to build a language model in which various word relationships can be integrated. We integrate two types of relationships, one extracted from WordNet and co-occurrence relationships having complementary advantages: the first is more precise and the second has a broader coverage. Our approach has been tested on several TREC collections and it achieves substantial and significant improvements with respect to the models without these relationships.