Emotional Chatting Machine: Emotional Conversation Generation with Internal and External Memory

By Hao Zhou, Minlie Huang, Tianyang Zhang, Xiaoyan Zhu, Bing-Qian Liu. 2017

808 citations recorded by the source corpus.

Perception and expression of emotion are key factors to the success of dialogue systems or conversational agents. However, this problem has not been studied in large-scale conversation generation so far. In this paper, we propose Emotional Chatting Machine (ECM) that can generate appropriate responses not only in content (relevant and grammatical) but also in emotion (emotionally consistent). To the best of our knowledge, this is the first work that addresses the emotion factor in large-scale conversation generation. ECM addresses the factor using three new mechanisms that respectively (1) models the high-level abstraction of emotion expressions by embedding emotion categories, (2) captures the change of implicit internal emotion states, and (3) uses explicit emotion expressions with an external emotion vocabulary. Experiments show that the proposed model can generate responses appropriate not only in content but also in emotion.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Lifelong and Continual Learning Dialogue Systems: Learning during Conversation (2021, background): “• Learning emotions and sentiments: Recognizing emotional state (Zhou et al. 2018a; Pamungkas 2019) and sentiments of the user and leverage it to generate empathetic responses can be useful to building therapeutic chatbots.”
Approximating Interactive Human Evaluation with Self-Play for Open-Domain Dialog Systems (2019, background): “Towards social and emotional coherence, previous works have investigated various features and loss functions based on emotion [32, 33, 13, 23].”
DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset (2017, methodology): “Existing work based on neural networks has examined their approaches on these datasets (Zhou et al., 2017; Wang et al., 2013; Sordoni et al., 2015; Serban et al., 2016b).”
Challenges in Building Intelligent Open-domain Dialog Systems (2019, background): “Ct can also encode other (non-content) contexts such as persona [103, 175, 186] for personalized dialog generation, emotion labels [4, 189] for emotional response generation, and knowledge graphs [36, 190] for knowledge-aware response generation.”
Style Transfer in Text: Exploration and Evaluation (2017, methodology): “(Zhou et al. 2017) controls emotion of conversation, it also uses a classifier to evaluate chatbot generated emotional response.”
A Survey on Dialogue Systems: Recent Advances and New Frontiers (2017, background): “[109] added emotion embedding into a generative model and achieved good performance in perplexity.”
Response Generation by Context-aware Prototype Editing (2018, methodology): “Apart from work on “safe response”, there is a growing body of literature on style transfer (Fu et al. 2018; Wang et al. 2017) and emotional response generation (Zhou et al. 2017).”
I Know the Feeling: Learning to Converse with Empathy (2018, background): “A few works focus on controlling the emotional content of a text response either through a manually speciﬁed target (Zhou & Wang, 2018; Zhou et al., 2017; Wang & Wan, 2018; Hu et al., 2017; Huang et al., 2018) or through a general term to encourage higher levels of affect (Asghar et al., 2018),…”
Towards Empathetic Open-domain Conversation Models: A New Benchmark and Dataset (2018, background): “…other works have focused on controlling the emotional content of a text response either through a manually speciﬁed target (Zhou and Wang, 2018; Zhou et al., 2018; Wang and Wan, 2018; Hu et al., 2017; Huang et al., 2018) or through a general term to encourage higher levels of affect (Asghar et…”
A Survey of Knowledge-enhanced Text Generation (2020, methodology): “Dialogue system [126, 132, 162, 179] and summarization [80, 81, 84] are two mainstream tasks that have adopted keyword knowledge.”
Dynamic Explainable Recommendation Based on Neural Attentive Models (2019, methodology): “GRU has been demonstrated to be effective in many neural language processing (NLP) tasks (Cho et al. 2014; Zhou et al. 2017; Noh, Hongsuck Seo, and Han 2016).”
A survey on empathetic dialogue systems (2020, methodology): “Since then, dialogue systems are either designed to perform specific tasks such as flight booking [6], healthcare [7], political debate [8], hence termed “taskspecific dialogue systems”, or to chitchat as a way of entertainment [9], hence called “chatbots”.”
Plug-and-Play Conversational Models (2020, background): “, 2020), emotions (Li et al., 2017; Rashkin et al., 2019; Zhou et al., 2018; Fan et al., 2020; Li et al., 2020), diversity (Li et al.”
Conversational Memory Network for Emotion Recognition in Dyadic Dialogue Videos (2018, background): “Previous research in affective computing has looked at di-alogues as an essential basis to learn emotional dynamics (Sidnell and Stivers, 2012; Poria et al., 2017a; Zhou et al., 2017).”
Knowledge Aware Emotion Recognition in Textual Conversations via Multi-Task Incremental Transformer (2020, background): “In addition, it can also be applied to chat robots to analyze the user’s emotional state in real time and generate emotion-aware responses (Poria et al., 2019b; Zhou et al., 2018a; Huang et al., 2018).”
The Rapidly Changing Landscape of Conversational Agents (2018, methodology): “Building on works like this the Emotional Chatting Machine model proposed by Zhou et al [31] is a model which generates responses that are not only grammatically consistent but also emotionally consistent.”
Beneath the Tip of the Iceberg: Current Challenges and New Directions in Sentiment Analysis Research (2020, background): “, 2017) or seq2seq models that are customized to generate emotionally-conditioned text (Zhou et al., 2018; Asghar et al., 2018).”
Adversarial Language Games for Advanced Natural Language Intelligence (2019, background): “…systems (also known as chatbots) interact with human in open domains naturally to provide entertainment, and typically generate responses by maximizing the likely-hood of human responses (Ritter et al., 2011; Banchs and Li, 2012; Li et al., 2016a,b; Serban et al., 2016, 2017; Zhou et al., 2018).”
Diversifying Dialogue Generation with Non-Conversational Text (2020, methodology): “As the standard seq2seq model trained only on D tends to generate over-generic responses, our purpose is to diversify the generated responses by leveraging the non-conversational corpus D T , which are semantically and syntactically much richer than responses contained in D .”