A Study of BERT for Non-Factoid Question-Answering under Passage Length Constraints
This work addresses passage length challenges in QA systems, but it is incremental as it applies existing BERT methods to a specific task.
The study tackled non-factoid question-answering by fine-tuning BERT for passage re-ranking under length constraints, achieving substantial improvements over state-of-the-art methods.
We study the use of BERT for non-factoid question-answering, focusing on the passage re-ranking task under varying passage lengths. To this end, we explore the fine-tuning of BERT in different learning-to-rank setups, comprising both point-wise and pair-wise methods, resulting in substantial improvements over the state-of-the-art. We then analyze the effectiveness of BERT for different passage lengths and suggest how to cope with large passages.