SPBERTQA: A Two-Stage Question Answering System Based on Sentence Transformers for Medical Texts

TL;DR

This paper introduces SPBERTQA, a two-stage question answering system for Vietnamese medical texts, utilizing Sentence-BERT and BM25, and presents a new healthcare QA dataset for Vietnamese.

Contribution

The paper develops a novel two-stage QA system based on Sentence-BERT and creates the first Vietnamese healthcare QA dataset with over 10,000 pairs.

Findings

The system outperforms traditional QA methods.

The dataset enables better evaluation of Vietnamese medical QA.

Experimental results show improved accuracy and relevance.

Abstract

Question answering (QA) systems have gained explosive attention in recent years. However, QA tasks in Vietnamese do not have many datasets. Significantly, there is mostly no dataset in the medical domain. Therefore, we built a Vietnamese Healthcare Question Answering dataset (ViHealthQA), including 10,015 question-answer passage pairs for this task, in which questions from health-interested users were asked on prestigious health websites and answers from highly qualified experts. This paper proposes a two-stage QA system based on Sentence-BERT (SBERT) using multiple negatives ranking (MNR) loss combined with BM25. Then, we conduct diverse experiments with many bag-of-words models to assess our system's performance. With the obtained results, this system achieves better performance than traditional methods.