# Emotionally-Aware Chatbots: A Survey

**Authors:** Endang Wahyu Pamungkas  
**arXiv:** [1906.09774](https://arxiv.org/abs/1906.09774)  
**Date:** 2019-06-24  
**Summary page:** https://www.summarizepaper.com/en/paper/1906.09774v1/

## Key points

- Development of emotionally-aware chatbots (EAC) and lack of research in this area
- Most EACs use a neural-based approach and incorporate emotion classifiers
- Recent studies propose new datasets for building EAC in various languages
- Evaluation criteria for emotionally-aware chatbots include content, emotion, consistency, logic, naturalness, empathy/sympathy
- Evaluation criteria measured using scales ranging from 0 to 2 or through human judgment
- Some studies use t-tests or inter-annotator agreement measurements to assess system performance
- Previous studies on chatbot development do not focus on emotional aspect
- References to works providing history of chatbot technology development and classification based on goals and techniques
- Table 2 summarizes available datasets for emotionally-aware chatbots
- Table 3 summarizes available affective resources for emotion classification tasks

## Summary

The existing summary discusses the development of emotionally-aware chatbots (EAC) and highlights the lack of research in this area. It also mentions that most EACs use a neural-based approach and incorporate emotion classifiers in their architecture. The summary further predicts that the development of EAC will continue to gain attention, as evidenced by recent studies proposing new datasets for building EAC in various languages. In addition to this information, the expanded summary provides details about the evaluation criteria used for emotionally-aware chatbots. It mentions different aspects such as content, emotion, consistency, logic, naturalness, and empathy/sympathy that are measured using scales ranging from 0 to 2 or through human judgment. Some studies use t-tests or inter-annotator agreement measurements to assess system performance. The related work section briefly reviews previous studies on chatbot development but notes that there is no specific study focusing on summarizing the development of chatbots with an emotional aspect. It references other works that provide a history of chatbot technology development and classify chatbots based on goals and development techniques. Furthermore, two tables are mentioned: Table 2 summarizes available datasets for emotionally-aware chatbots while Table 3 summarizes available affective resources for emotion classification tasks. These tables provide additional information about the resources and tools used in developing emotionally-aware chatbots. Overall, the refined detailed longer summary provides a comprehensive overview of emotionally-aware chatbot development including evaluation criteria, related work and available resources in this field which can be used to build more advanced systems with greater accuracy and reliability.