Recently, due to the availability of vast information, which results more difficult to find out and discover what we need. We need tools which help us to establish, examine and recognize these huge quantities of information. For automatically establishing, understanding, examining, and summarizing large automatic accounts, subject modeling delivers some processes: 1. Determine the unseen subjects in the collection 2. Make notes on the documents giving to these subjects 3. Use comments to establish, review and examine. Hence for these purpose Probabilistic Latent Semantic Indexing approach is used to automate document indexing by using a statistical latent class model for factor analysis of count data. In this paper, we find out a set of query-focused which are summarizing from search results. As there may be several subjects related to a given query in the search results, hence to summarize these results in proper order, they should be classified into subjects, and then every subject should be summarized separately. There are two types of redundancies need to be reduced in this summarization process. First, every subject summary should not comprise any redundancy. Second, a subject summary should not be analogous to any other subject summary. In the summarization process, we emphasis on the document grouping process as well as reducing the redundancy between summaries. In this paper, we also suggest the PLSI approach which is a way to summarize the search results. Due to the process of evaluation results, our method accomplishes well in categorizing search results and reducing the redundancy between summaries.