Automatic Dialogue Generation with Expressed Emotions

Abstract

Despite myriad efforts in the literature designing neural dialogue generation systems in recent years, very few consider putting restrictions on the response itself. They learn from collections of past responses and generate one based on a given utterance without considering, speech act, desired style or emotion to be expressed. In this research, we address the problem of forcing the dialogue generation to express emotion. We present three models that either concatenate the desired emotion with the source input during the learning, or push the emotion in the decoder. The results, evaluated with an emotion tagger, are encouraging with all three models, but present better outcome and promise with our model that adds the emotion vector in the decoder.

ANA at SemEval-2019 Task 3: Contextual Emotion detection in Conversations through hierarchical LSTMs and BERT

“…Although its potential convenience, detecting emotion in textual conversation has seen limited attention so far. One of the main challenges is that one users utterance may be insufficient to recognize the emotion (Huang et al, 2018). The need to consider the context of the conversion is essential in this case, even for human, specifically given the lack of voice modulation and facial expressions.…”

Generating Responses Expressing Emotion in an Open-Domain Dialogue System

“…The work in [10] is able to generate personalized responses given a specific speaker, which can be considered as one of the first attempts that control the generations of seq2seq models. In terms of controlling emotions, [32] tackles this problem with a sophisticated memory mechanism while [7] uses three concise but efficient models to achieve equally good performance.…”

“…We propose three models (Enc-bef, Enc-aft and Dec) in [7]. Enc-bef and Enc-aft are models that inject an emotion e in the encoder by putting special tokens before or after the input sequence X.…”

“…The Encoder-Decoder framework provides us with a flexible foundation which makes joining additional modules into it straightforward and intuitive. In this work, inspired by [7] and [32], four models are proposed.…”

“…Dec-start In [7] Enc-bef and Enc-aft models have been shown to be successful and effective. By creating a special token T e for every emotion, these two methods are essentially modifying X to T e ; X or X; T e in both training and evaluating.…”