A fuzzy ranking approach for improving search results in Turkish as an agglutinative language

Abstract

This study proposes a fuzzy ranking approach, designed for Turkish as an agglutinative language, that focuses on improving stemming techniques via using distances of characters in its search algorithm. Various studies focused on search engines are based on using stemming techniques in indexing process because of the higher percentage of relevancy that these techniques provide. However, stemming techniques may have negative effects on search results in some queries. While analyzing the search results to find the query terms those give irrelevant results and why, we observe that user's query suffixes are crucial in search performance. Therefore, the proposed fuzzy ranking approach supports traditional stemming approaches with the use of suffixes. The search results of this approach are significantly better than stemming techniques in where stemming technique is ineffective. In terms of overall results, the fuzzy ranking approach also gives satisfactory results when compared with stemming techniques such as a Turkish stemmer (19.43% of improvement) and word truncation technique (12.61% of improvement). Moreover, it is statistically better than no stemming with 28.61% of improvement. (C) 2011 Elsevier Ltd. All rights reserved.