publications

In the realm of human-AI dialogue, the facilitation of empathetic responses is important. Validation is one of the key communication techniques in psychology, which entails recognizing, understanding, and acknowledging others' emotional states, thoughts, and actions. This study introduces the first framework designed to engender empathetic dialogue with validating responses. Our approach incorporates a tripartite module system: 1) validation timing detection, 2) users' emotional state identification, and 3) validating response generation. Utilizing Japanese EmpatheticDialogues dataset - a textual-based dialogue dataset consisting of 8 emotional categories from Plutchik's wheel of emotions - the Task Adaptive Pre-Training (TAPT) BERT-based model outperforms both random baseline and the ChatGPT performance, in term of F1-score, in all modules. Further validation of our model's efficacy is confirmed in its application to the TUT Emotional Storytelling Corpus (TESC), a speech-based dialogue dataset, by surpassing both random baseline and the ChatGPT. This consistent performance across both textual and speech-based dialogues underscores the effectiveness of our framework in fostering empathetic human-AI communication.

Validation, a key communication technique that entails recognizing, understanding, and accepting others' emotional states, thoughts, and actions, plays a vital role in fostering strong interpersonal relationships. This work introduces a pioneering system for generating validating responses to foster empathetic dialogue. Utilizing the TUT Emotional Storytelling Corpus (TESC) - a Japanese multi-turn corpus detailing 8 emotional categories from Plutchik's wheel of emotions - our Task Adaptive Pre-Training (TAPT) BERT-based approach achieves 56.5% precision and accuracy in pinpointing validating response generation moments, and 88.6% precision and accuracy in emotion classification during validating response generation. For practical evaluation, we compared our system against Seq2Seq and ChatGPT empathetic responses through human assessment.

Empathy is the capacity to place oneself in another person’s position. To show empathy in conversation, validation is one of the methods that can be used. Validation is a technique to show understanding of what has happened and how they feel. This study aims to develop a model that determines the necessity of generating validating responses in conversation through its prior utterance. TUT emotional storytelling corpus (TESC) has been used in this study. The prior utterances have been analyzed for various aspects, and the results indicate that emotional phrases, laughing, emphasizing phrases, and particle types in the last word affect the generation of validating responses and thus serve as the basis for developing the model in this study. Using a logistic regression model, we can predict the validating responses with an accuracy of 62.7%, and a recall of 60.0%.