Sentence Extraction-Based Machine Reading Comprehension for Vietnamese

TL;DR

This paper introduces UIT-ViWikiQA, a new Vietnamese dataset for sentence extraction-based machine reading comprehension, along with models and analysis to advance NLP research in Vietnamese.

Contribution

It presents the first Vietnamese dataset for sentence extraction MRC, a conversion algorithm, and evaluates models including XLM-R_Large on this dataset.

Findings

XLM-R_Large achieves 85.97% EM and 88.77% F1 score.

Analysis of question types and context effects on model performance.

Highlights challenges in Vietnamese MRC from the dataset.

Abstract

The development of natural language processing (NLP) in general and machine reading comprehension in particular has attracted the great attention of the research community. In recent years, there are a few datasets for machine reading comprehension tasks in Vietnamese with large sizes, such as UIT-ViQuAD and UIT-ViNewsQA. However, the datasets are not diverse in answers to serve the research. In this paper, we introduce UIT-ViWikiQA, the first dataset for evaluating sentence extraction-based machine reading comprehension in the Vietnamese language. The UIT-ViWikiQA dataset is converted from the UIT-ViQuAD dataset, consisting of comprises 23.074 question-answers based on 5.109 passages of 174 Wikipedia Vietnamese articles. We propose a conversion algorithm to create the dataset for sentence extraction-based machine reading comprehension and three types of approaches for sentence extraction-based machine reading comprehension in Vietnamese. Our experiments show that the best machine model is XLM-R_Large, which achieves an exact match (EM) of 85.97% and an F1-score of 88.77% on our dataset. Besides, we analyze experimental results in terms of the question type in Vietnamese and the effect of context on the performance of the MRC models, thereby showing the challenges from the UIT-ViWikiQA dataset that we propose to the language processing community.