Flexible Semantic Qur’an Question Answering Using Graph-Based Summarization and KNN

Researchers in the computer science field have been attracted by Qur'an-based research. This research area focuses on representing the ontology-based Qur'an. A semantic-based search will be beneficial in extracting information from the Quran, which has complex knowledge and language. This work aims to develop flexible semantic Qur'an question-answering by applying graph-based summarization and K-nearest neighbors (KNN) methods to add flexibility to semantic-based searches in the Indonesian Language. Meanwhile, the Qur'an is based on a unique Arabic language. It is a part of the complexity of this work. The graph-based summarization method effectively summarizes a complex question. It was proved by ROUGE testing with F1, precision, and recall scores of 72%, 62%, and 72%, respectively. The KNN method evaluated by the expert resulted in an average approval percentage on the 1st, second, and third topics of 62.11%, 66.15%, and 19.61%. As for other issues related to the questions, 70% needs to be displayed. The analysis of the obtained result indicates that the classification step needs to be improved in the tiny dataset. This work will contribute to Qur'an Question Answering as it is considered that the Qur'an is a different object compared to the other content of Question Answering. The Qur'an is an object that contains a massive amount of multi-interpretation. Lots more work in the future. The dataset is also limited by the scope of the theme of this research, which is only the pillars of Islam, so many topics still need to be included in the dataset