Arabic Query Reformulation using Harmony Search Algorithm

Abstract

The search on the web helps users to find the most relevant information. However, this is not always verified because many users search with ambiguous queries to find information. The ambiguous queries can return irrelevant results. Query reformulation aims to get more relevant information to query despite the ambiguity in the query. It works to disambiguate the meaning of a word in a query and finds its alternative. This is a big challenge in Arabic queries. In addition, there is another challenge which is the delay in the online word preprocessing. The reducing of delay should be considered along with the increasing of the accuracy. This work takes into account these two challenges by using a powerful met heuristics search algorithm that is called harmony search. The algorithm is used to reformulate a query with high accuracy and low delay. The proposed system that uses this algorithm is tested on a set of ambiguous Arabic queries and two different Arabic corpora called ZAD and NLEL. The system is able to get more relevant documents to a user query. Experimental results show significant superiority of the proposed query reformulation system in terms of response time, precision, recall, and mean average precision. The proposed system outperforms the baseline system in a rate up to (33.22 %) and reduces the latency in an approximate rate up to (0.064) second for each query.