# DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset

**Authors:** Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, Shuzi Niu  
**arXiv:** [1710.03957](https://arxiv.org/abs/1710.03957)  
**Date:** 2017-10-11  
**Summary page:** https://www.summarizepaper.com/en/paper/1710.03957v1/

## Key points

- Authors developed a high-quality multi-turn dialog dataset called DailyDialog
- Dataset features human-written language and accurately reflects daily communication patterns
- Covers a wide range of topics related to everyday life
- Dataset manually labeled with information about communication intention and emotion
- Existing approaches evaluated on the DailyDialog dataset to provide benchmark for performance evaluation in dialog systems research field
- Study contributes to natural language processing and dialog systems field by providing researchers with a high-quality resource and insights into human communication patterns and emotions in various contexts.

## Summary

The authors Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao and Shuzi Niu have developed a high-quality multi-turn dialog dataset called DailyDialog. This dataset is unique in that it features human-written language which is less noisy than other datasets. It accurately reflects our daily communication patterns and covers a wide range of topics related to everyday life. To further enhance the dataset, the authors manually labeled it with information about communication intention and emotion. The authors also evaluated existing approaches on the DailyDialog dataset in order to contribute to the research field of dialog systems by providing a benchmark for performance evaluation. Through these evaluations they hope to facilitate advancements in dialog system technology. Overall this study presents an important contribution to the field of natural language processing and dialog systems as the DailyDialog dataset not only provides researchers with a high-quality resource but also offers insights into human communication patterns and emotions in various contexts.