MISC: A Mixed Strategy-Aware Model integrating COMET for Emotional Support Conversation

Abstract

Applying existing methods to emotional support conversation-which provides valuable assistance to people who are in need-has two major limitations: (a) they generally employ a conversation-level emotion label, which is too coarse-grained to capture user's instant mental state; (b) most of them focus on expressing empathy in the response(s) rather than gradually reducing user's distress. To address the problems, we propose a novel model MISC, which firstly infers the user's fine-grained emotional status, and then responds skillfully using a mixture of strategy. Experimental results on the benchmark dataset demonstrate the effectiveness of our method and reveal the benefits of fine-grained emotion understanding as well as mixed-up strategy modeling. Our code and data could be found in https: //github.com/morecry/MISC.

FADO: Feedback-Aware Double COntrolling Network for Emotional Support Conversation

“…Recently, emotional support conversation has attracted much research interest [11,28,17,33]. In general, emotional support conversation models can be divided into two categories, including explicitly modeling psychological factors at cognitive level with the graph network [17] and utilizing the strategy [11,28] for generating supportive responses.…”

“…Recently, emotional support conversation has attracted much research interest [11,28,17,33]. In general, emotional support conversation models can be divided into two categories, including explicitly modeling psychological factors at cognitive level with the graph network [17] and utilizing the strategy [11,28] for generating supportive responses. For instance, [17] propose a global-to-local hierarchical graph network to capture the psychological features and model hierarchical relationships between them to generate supportive response.…”

“…As for the startegy-aware generation, [11] use a special token to represent each strategy and append the strategy token before the response to make a generation. [28] design a mixed strategy-aware model integrating COMET [1], a pre-trained commonsense language model, to respond skillfully. However, they ignore dual-level feedbacks and simply optimize the strategy with the ground truth, regardless of considering encouraging/penalizing strategies that obtain positve/negative feedbacks.…”

“…During the emotional support scenery, researches [5,11,28] have shown the great significance for dialog agents to select a proper supportive strategy, which guides the response generation and thus comforts the help-seekers' emotion. However, there are some issues in the strategy selection and supportive response generation.…”

“…Firstly, as shown in Fig. (a), mainstream studies [11,28] simply leverage the dialog history to select a strategy and optimize it with the ground truth, ignoring the dual-level feedback of the help-seeker (as shown in red in Fig. 1 (b)), which leads to the wrong and user-irrelevant (inability to understand the user's feelings) strategy prediction.…”