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

TL;DR

This work introduces the Emotional Chatting Machine (ECM), a neural framework that integrates emotion into large-scale open-domain dialogue generation. It combines three mechanisms—emotion category embeddings, an internal memory that tracks dynamic emotional state, and an external memory that explicitly selects emotion words—to produce responses that are both coherent and emotionally aligned with a given target emotion. Through a large emotion-annotated dataset (ESTC) and comprehensive automatic and manual evaluations, ECM outperforms a standard Seq2Seq and an emotion-embedding baseline in emotion accuracy and, in many cases, content quality. The results demonstrate that explicit modeling of emotional expression can meaningfully enhance user-facing conversational agents, with pathways for future work in automatic emotion selection and richer emotion interactions.

Abstract

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.

Figures (5)

• Figure 1: Overview of ECM (the grey unit). The pink units are used to model emotion factors in the framework.
• Figure 2: Data flow of the decoder with an internal memory. The internal memory $\bm{M}^I_{e,t}$ is read with the read gate $\bm{g}^r_t$ by an amount $\bm{M}^I_{r,t}$ to update the decoder's state, and the memory is updated to $\bm{M}^I_{e,t+1}$ with the write gate $\bm{g}^w_t$.
• Figure 3: Data flow of the decoder with an external memory. The final decoding probability is weighted between the emotion softmax and the generic softmax, where the weight is computed by the type selector.
• Figure 4: Sample responses generated by Seq2Seq and ECM (original Chinese and English translation, the colored words are the emotion words corresponding to the given emotion category). The corresponding posts did not appear in the training set.
• Figure 5: Visualization of emotion interaction.