Soft Clustering for Optimal Topic Based Document Retrieval

Abstract

Topic modeling is a type of the statistical model that has been proven successful for discovering topics from document. The extracted topics have been used as alternative indexes to overcome the drawbacks of the standard word based document retrieval. The purpose of this paper is to examine the use of clustering methods for the topic-based document retrieval. This leads us to propose a soft clustering method that can be used for optimizing the performance of the topic-based document retrieval. We show that by applying the soft clustering method where a document is assigned to several significant clusters, we can improve the performance of hard clustering method while still reducing the number of considered documents to a reasonable size.