هاله عباس

The development and eva‎luation of question-answering (QA) systems are essential for advancing natural language processing (NLP) technologies and improving human-computer interaction. This study focuses on creating and assessing an Islamic question answering system using English-language question-answer pairs from a specific website. The research began with a thorough preprocessing of the textual data. After preprocessing, the data was embedded using the SBERT model to enable effective machine learning. A novel model based on deep clustering was then proposed. This model combines elements of deep neural networks (DNNs) with k-means clustering. It uses fully connected layers with the rectified linear unit (ReLU) activation function to enhance learning capabilities and feature representation. Experimental results showed that the system achieved a BLEU score of 0.260. Additionally, the system performed well in terms of ROUGE scores, achieving ROUGE1-mid of 0.380, ROUGE2-mid of 0.168, and ROUGEL-mid of 0.235. These results highlight the effectiveness of the proposed system in accurately clustering question-answer pairs and identifying relevant answers.

پردازش زبان طبيعي , سيستم پرسش و پاسخ

Natural Language Processing , Question Answering System

Hale Abbas

Dr. Behrooz Minaei