Information retrieval in Telugu language using Synset relationships

Abstract

Information Technology brought many applications of Information Retrieval as simple as possible through Web and other Digital Information Access Environment. There is a demand in building Applications of IR in Local languages, which allows common people with minimal knowledge in at least one language to avail the information services. Word mismatch is a common problem in IR System Applications. The research is started to overcome this problem in 4 decades back and still the problem is unsolved. The methods were moving from word replacement to knowledge replacement to solve word mismatch and concept mismatches. This paper mainly explored towards solving word mismatch problem in India language context particularly Telugu language. IR in Telugu Language is in inception phase and need huge research to solve many issues in information processing and retrieval. Cross Lingual IR is quite comfortable stage with parallel rule conversion techniques, where as it is difficult to process Telugu alone and build Information Retrieval System. Telugu is morphologically rich with high conflational rate. In this paper we build IR in Telugu Language and extended to solve vocabulary mismatch problem using Synset relationships. When compared to base Retrieval system, expanded search yield offered results. This paper identified many issues related to topical search and given direction to solve with base proof with implementation results. We found improvement in Recall and precision with Expanded Search with Non-Expanded Retrieval.