Implicit Semantic Relations Identification through Distributed Representations for Effective Text Retrieval

Abstract

The modern commercial search methods have proved the efficiency of information retrieval (IR) technologies and made knowledge dissemination simpler where as finding the relevant text documents, given a query, is becoming much more complex day after day. Almost all information retrieval systems struggle to retrieve relevant entries on the top of ranked list of documents. Users are interested to retrieve the content satisfying their information needs rather than just retrieving the documents having the query terms supplied by them. For this task, traditional information retrieval methods hardly yield the results having higher-order relations between the given query and the documents in the collection. So, it becomes necessary to find and encode the higher-order feature associations present in text documents. This paper discusses the use of a distributed representation, namely random indexing for an effective retrieval of relevant text documents. This type of distributed representation would be scalable with modern computing facilities and flexible to develop knowledge-based applications, which may require the process of identifying implicit semantic relations though higher-order term associations.

http://dx.doi.org/10.14429/djlit.31.4.1104