JPS-daprinfo: A Dataset for Japanese Dialog Act Analysis and People-related Information Detection

TL;DR

This paper introduces JPS-daprinfo, a Japanese dialogue dataset with annotated labels for dialog act analysis and people-related information detection, based on 50 interview dialogues totaling over 30 hours.

Contribution

It provides a new annotated Japanese dialogue dataset with 13 labels, specifically designed for dialog act analysis and information detection tasks.

Findings

Annotated 20,130 sentences with 13 labels

Dataset based on 50 native Japanese interview dialogues

Facilitates research in Japanese dialog analysis

Abstract

We conducted a labeling work on a spoken Japanese dataset (I-JAS) for the text classification, which contains 50 interview dialogues of two-way Japanese conversation that discuss the participants' past present and future. Each dialogue is 30 minutes long. From this dataset, we selected the interview dialogues of native Japanese speakers as the samples. Given the dataset, we annotated sentences with 13 labels. The labeling work was conducted by native Japanese speakers who have experiences with data annotation. The total amount of the annotated samples is 20130.