Modelling Context Emotions using Multi-task Learning for Emotion Controlled Dialog Generation

Abstract

A recent topic of research in natural language generation has been the development of automatic response generation modules that can automatically respond to a user's utterance in an empathetic manner. Previous research has tackled this task using neural generative methods by augmenting emotion classes with the input sequences. However, the outputs by these models may be inconsistent. We employ multitask learning to predict the emotion label and to generate a viable response for a given utterance using a common encoder with multiple decoders. Our proposed encoder-decoder model consists of a self-attention based encoder and a decoder with dot product attention mechanism to generate response with a specified emotion. We use the focal loss to handle imbalanced data distribution, and utilize the consistency loss to allow coherent decoding by the decoders. Human evaluation reveals that our model produces more emotionally pertinent responses. In addition, our model outperforms multiple strong baselines on automatic evaluation measures such as F1 and BLEU scores, thus resulting in more fluent and adequate responses.

Untitled

“…The study provides a systematic review of approaches in building an emotionally-aware chatbot (EAC). In [55] the authors use multi-task learning to predict the emotion label and generate a valid response for a given utterance. Their model consists of a self-attention based encoder and a decoder with dot product attention mechanism to generate response with a specified emotion and produce more emotionally relevant responses.…”

AI-Based Conversational Agents: A Scoping Review From Technologies to Future Directions

“…It uses an attention weight to measure the importance of each word in the input text. Different attention mechanisms like self-attention [38], multi head-attention [39], word-level attention [40], hierarchical attention [41], are used in various research papers in conversational agents. Deep learning methods such as RNN and its different variations has own limits.…”

“…So, ambiguity in word semantic interpretations is one of the important issues in CAs. [38] -Datasets plays an important role as training data is required to understand intent and context and respond naturally to user. There are many challenges related to datasets like small datasets available, scarcity of labeled data, unbalanced distribution of data, less variety of labels in datasets, and lack of representative publicly available datasets.…”

“…First is relatedness, that is how different tasks are related to each other and kinds of learning tasks such as supervised like classification or regression tasks, unsupervised learning, clustering task, and many more learning tasks like semisupervised, reinforcement learning. In [38] proposed, a model where authors used multi-task learning to classify the emotions and to generate the response based on a given sentence with a common encoder and multiple decoders.…”