Summary-Based retrieval is based on the hypothesis that terms in summary should be more important than other terms not in summary. Recent developments in the language modeling approach to information retrieval have motivated the study of this problem within this new retrieval framework. In the proposed research, two approaches to summary-based retrieval, namely ranking documents directly (SQL) and smoothing documents with summaries (SBDM) are investigated. Results on TREC collections show that, with the proposed models, summary-based retrieval models can perform consistently across collections and significant improvements over document-based retrieval can be obtained. There are two main contributions in this paper. On the one hand, summarization method of retrieval-oriented is examed and effect of this method on information retrieval. On the other hand, the new retrieval model for summary-based information retrieval models is proposed.