An Ontology Based Approach to Enhance Information Retrieval from Al-Shamelah Digital Library

  • Mohammed G. Al-Masri -----> Dr. Iyad Al-Agha

With the huge number of Islamic references that emerged over hundreds of years, several difficulties were introduced when searching this huge Heritage. This has challenged researchers to develop computer systems to facilitate information retrieval and extraction from the Islamic heritage. One of these systems is Al-Shamelah Digital Library (ADL), Al-maktabah Al-Shamelah, which is a huge database containing thousands of books in different Islamic disciplines.
Information retrieval from ADL is mainly based on keyword matching, and does not provide semantic interpretations of Islamic texts. Current information retrieval capabilities in ADL do not provide the possibility to handle complex queries, provide intelligent results based on the semantic processing of content, or extract implicit relations and meanings from text. Islamic texts, such as the Quran and Sunna, are rich sources of knowledge with many underlying thoughts, meanings and laws. They often include many metaphors and figurative speech which cannot be directly interpreted. In addition, the Islamic Knowledge is known to be very diverse and interrelated.
Traditional search services provided by ADL only allows access to information without revealing the relationships or dependencies between various information resources.
Driven by the above challenges, this thesis proposes a system which called OntoADL that supports semantic search and recommendation over a subset of ADL. At the core of OntoADL is our ontology-based approach that leverages ontology-based annotations to produce highly relevant search results and to offer recommendations of related topics of interest. It is based on the Hadith ontology that we built, in cooperation with domain experts, to model the different entities included in the Prophetic Medicine, the domain we selected to assess our approach. This thesis explains the design and architecture of OntoADL, focusing on how ontology-based reasoning can result in intelligent results that meets the user’s interests. The search service in OntoADL was evaluated by being compared with the search service in the conventional ADL. The OntoADL achieved (90%) recall and (83%) precision while the ADL system achieved 69% recall and 30% precision.
These results indicate that our approach surpasses the ADL search facility. To assess the rankings of retrieved results, the Mean Average Precision (MAP) was calculated for both OntoADL and ADL. OntoADL achieved 86% MAP while the ADL system achieved 36%. This result also indicates that OntoADL outperforms the conventional ADL in terms of the ranked retrieval results.