Automatic Query Expansion for Arabic Information Retrieval

Abstract

Query expansion (QE) is a successful idea to overcome the weaknesses in the information retrieval performance. The QE requires finding out appropriate word synonyms of the query words in a process that can be made automatically without any user intervention. The candidate synonyms should be associated with an accurate meaning (sense) of the original word. Arabic language is rich in multiple meanings and this requires using the so-called word sense disambiguation (WSD). WSD in general is a task to discover the correct sense of a word within context. To disambiguate the word sense, three different traditional semantic measures are tested in this work; they are called lch, wup, and path respectively. The proposed system uses these measures along with an automatic synonym selection method employed to expand the query. The proposed system outperforms the traditional baseline system that has no query expansion technique in a rate from 10% to 18 % and reduces the latency in an approximate rate from 0.232 to 0.283 second for each query.