[Paper Review] KorQuAD1.0: Korean QA Dataset for Machine Reading Comprehension

Seungyoung Lim, Myungji Kim+1 more|arXiv (Cornell University)|Sep 16, 2019

Topic Modeling7 references37 citations

TL;DR

KorQuAD1.0 introduces a large-scale Korean extractive QA dataset with 70,000+ human-generated question–answer pairs on Korean Wikipedia to advance MRC research.

ABSTRACT

Machine Reading Comprehension (MRC) is a task that requires machine to understand natural language and answer questions by reading a document. It is the core of automatic response technology such as chatbots and automatized customer supporting systems. We present Korean Question Answering Dataset(KorQuAD), a large-scale Korean dataset for extractive machine reading comprehension task. It consists of 70,000+ human generated question-answer pairs on Korean Wikipedia articles. We release KorQuAD1.0 and launch a challenge at https://KorQuAD.github.io to encourage the development of multilingual natural language processing research.

Motivation & Objective

• Motivate and enable robust machine reading comprehension research in Korean.
• Provide a large-scale, human-generated extractive QA dataset for Korean to support evaluation and model development.
• Promote multilingual NLP research by releasing KorQuAD1.0 and hosting a challenge.

Proposed method

• Construct a large-scale Korean QA dataset with 70,000+ human-generated question–answer pairs.
• Format the task as extractive machine reading comprehension on Korean Wikipedia articles.
• Release KorQuAD1.0 publicly and launch a challenge to foster NLP research.

Experimental results

Research questions

• RQ1 How well can models perform extractive QA on Korean Wikipedia passages?
• RQ2 Can the dataset facilitate advancements in multilingual NLP research and evaluation for Korean?
• RQ3 What benchmarks and baselines emerge from KorQuAD1.0 as a resource for MRC in Korean?

Key findings

• The dataset contains 70,000+ human-generated question–answer pairs on Korean Wikipedia articles.
• KorQuAD1.0 is released to support and motivate multilingual NLP research in QA tasks.
• A challenge following the dataset release is launched to encourage development of MRC models for Korean.