Arabic Question Answering Using Ontology

Currently, with the massive amount of data that is being posted on the web at a rapid pace, users regularly have inquiries and they expect to find out short and precise answers. Semantic web and ontology technologies are becoming the essential components to represent domain-specific data that can be utilized in Question Answering System (QAS). In this paper, we introduce an Arabic QAS based on the domain knowledge or ontology in order to answer natural language inquiries. Prior to the implementation, it was crucial to perform some Natural Language Processing (NLP) tasks that assisted in analyzing the questions such as normalization, tokenizing, removing the stop words, stemming and tagging. Furthermore, we present how to develop the ontology through the Protégé tool, how to translate the inquiries into triple patterns and build the SPARQL queries which are the mechanism to retrieve the answer from Resource Description Framework (RDF) data. As far as Arabic is concerned, Arabic Question Answering (QA) is still limited and did not reach the similar level of English QA because of challenges on the Arabic language, for example, complexity in morphology derivational and inflectional, and words suffer from the scare of vowels. In addition, Arabic language orthography does not use capital letters or the like, which affect on Named Entity Recognitions. Hence, it is an opportunity to focus on Arabic QA using ontology toward getting a clear concept of this semantic-based approach. The result of the experimental results demonstrates the feasibility of constructing a QAS based on ontology. The proposed model has achieved promising results with accuracy of 81%, which provides an important indication for further in-depth study and analysis.