Improving Ranking-based Question Answering with Weak Supervision for Low-Resource Qur’anic Texts

Download paper here

🌟 The paper is titled: Improving Ranking-Based Question Answering with Weak Supervision for Low-Resource Qur’anic Texts. This work introduces, among other findings, a novel weakly supervised learning mechanism tailored for ranking-based question answering over the Holy Qur’an. Here are the main contributions of the research:

1. We present a novel learning method for machine reading comprehension that utilizes weak supervision signals specifically for question answering in a ranking context. This approach aligns closely with the inherent ranking nature of the task.
2. Using our proposed learning method, we achieve statistically significant, state-of-the-art performance on the QRCD dataset, surpassing results reported in prior studies.
3. To support further research, we’re making our trained models, experiments, and codes open-source for the community.

This paper is part of my series on the Qur’an QA challenge, organized by the BigIR group. (Dr. Tamer Elsayed, Dr. Rana Malhas, and Watheq Mansour) Qur’an QA is a research challenge that took place twice in 2023 and 2022 attracting young researchers to develop MRC and ad hoc search solutions for the Holy text.

Recommended citation:

Recommended citation:
Download Paper