Sistem Pencarian Ayat Al-Quran Berbasis Kemiripan Fonetis

Abstract

Searching Arabic text in the Holy Quran is not easy for users that do not have sufficient knowledge about Arabic language and script. Therefore, phonetic search can be used to facilitate users to search in the Holy Quran with their pronunciation represented in Latin script using regular alphabets. This research aims to build such search system, specifically for speakers of Bahasa Indonesia. A phonetic coding method regarding Quran recitation rules (tajweed) is proposed to match between Quran texts in Arabic script and user's queries in Latin script. Indexed trigram is used for approximate string matching. The system uses two search schemes: vowelized and non-vowelized search; two ranking methods: trigram count and trigram position ranking; and two search purposes: pronunciation and topic search. Experiment using user-generated queries and given relevance judgments shows that vowelized search performs better than non-vowelized search with 0.651 average precision (AVP) at the cost of longer search time. Trigram count ranking performs better than trigram position ranking with 0.668 AVP, although trigram position ranking can obtain higher precision at lower recall level. Pronunciation search queries perform better than topic search queries with 0.751 AVP.