[Paper Review] Emotionally-Aware Chatbots: A Survey

Endang Wahyu Pamungkas|arXiv (Cornell University)|Jun 24, 2019

Sentiment Analysis and Opinion Mining53 references35 citations

TL;DR

A systematic survey of emotionally-aware chatbots (EAC), tracing history from rule-based to neural approaches, outlining architectures, datasets, resources, evaluation methods, and future directions.

ABSTRACT

Textual conversational agent or chatbots' development gather tremendous traction from both academia and industries in recent years. Nowadays, chatbots are widely used as an agent to communicate with a human in some services such as booking assistant, customer service, and also a personal partner. The biggest challenge in building chatbot is to build a humanizing machine to improve user engagement. Some studies show that emotion is an important aspect to humanize machine, including chatbot. In this paper, we will provide a systematic review of approaches in building an emotionally-aware chatbot (EAC). As far as our knowledge, there is still no work focusing on this area. We propose three research question regarding EAC studies. We start with the history and evolution of EAC, then several approaches to build EAC by previous studies, and some available resources in building EAC. Based on our investigation, we found that in the early development, EAC exploits a simple rule-based approach while now most of EAC use neural-based approach. We also notice that most of EAC contain emotion classifier in their architecture, which utilize several available affective resources. We also predict that the development of EAC will continue to gain more and more attention from scholars, noted by some recent studies propose new datasets for building EAC in various languages.

Motivation & Objective

• Explain the motivation for humanizing chatbots through emotion to boost user engagement.
• Survey historical evolution and key approaches used to build emotionally-aware chatbots.
• Identify available datasets and affective resources for emotion understanding in chatbots.
• Review evaluation methodologies (qualitative and quantitative) used to assess EAC performance.
• Discuss challenges and predict future directions, including multilingual and contextual improvements.

Proposed method

• Review and synthesize prior work on EAC from rule-based beginnings to neural-based models.
• Summarize architectural patterns, emphasizing encoder-decoder seq2seq models with emotion signals.
• Catalog available emotion datasets and affective lexical resources, with language focus notes (English/Chinese).
• Describe emotion classifiers integrated into EAC and their role in response generation.
• Outline evaluation frameworks, including ISO 9241 aspects and automatic/manual metrics used in EAC studies.
• Highlight trends and future directions in EAC research.

Experimental results

Research questions

• RQ1 RQ1 How to incorporate emotion information in building an emotionally-aware chatbot?
• RQ2 RQ2 What are available resources that can be used in building emotionally-aware chatbots?
• RQ3 RQ3 How to evaluate the performance of emotionally-aware chatbots?

Key findings

• Emotionality has shifted from rule-based to neural-based approaches, with encoder-decoder architectures becoming dominant.
• Most emotionally-aware chatbots incorporate an emotion classifier to detect user emotions for guiding responses.
• Available datasets for EAC are predominantly English or Chinese and sourced from social media, online content, or crowdsourcing.
• Affective resources (lexicons) such as LIWC, ANEW, DepecheMood, and EmoWordNet are widely used for emotion classification.
• Evaluation of EAC uses qualitative (usability, ISO 9241) and quantitative methods (automatic metrics like perplexity, precision/recall, BLEU, and manual human judgments).
• Future work is expected to expand multilingual coverage and develop context-aware and more nuanced affective capabilities.