Japanese Multiple-Choice Machine Reading Comprehension Dataset JLPT-MC

To verify the effectiveness of the model, the author constructed a Japanese MCRC dataset based on the Japanese Language Proficiency Test (JLPT). This dataset covers the questions in the "Solution" section of JLPT's five test levels from 1992 to 2023, covering categories such as history, daily life, music culture, etc. This dataset extracts a total of 1731 pairs of context question candidate option triplets, including 1146 contexts, 1731 questions, and 6924 candidate options. To ensure that the model can be trained on pure text data, this article excludes non pure text data such as image selection, route planning, and graphic judgment. The processed triplet data consists of a total of 1329 pairs, including 904 contexts, 1329 questions, and 5316 candidate options. The dataset is named JLPT-MC.