Semantic Similarity Between Arabic Questions Using Support Vector Machines and Hungarian Method

Semantic tasks in Natural Language Processing (NLP), such as Semantic Textual Similarity and Textual Entailment, play a crucial role in numerous advanced NLP applications. Among these tasks, Semantic Question Similarity is particularly important for improving the performance of systems in areas such as Question Answering, chatbots, automated customer support, and Community Question Answering platforms. In this work, we propose an approach to deal with Semantic Question Similarity using the Hungarian method and Support Vector Machines. Our methodology builds upon an existing system originally designed for Recognizing Textual Entailment in Arabic, incorporating improvements and specific adaptations to address the particular characteristics of the Question Similarity task. The evaluation of our approach was performed on an existing dataset and has yielded encouraging results.