Towards Detecting Need for Empathetic Response in Motivational Interviewing

摘要: Empathetic response from the therapist is key to success of clinical psychotherapy, especially motivational interviewing. Previous work on computational modelling empathy in interviewing has focused offline, session-level assessment empathy, where captures all efforts that makes understand client's perspective and convey understanding client. In this position paper, we propose a novel task turn-level detection client need for empathy. Concretely, leverage pre-trained language models empathy-related general conversation corpora unique labeller-detector framework, labeller automatically annotates corpus with labels train detector determines We also lay out our strategies extending additional-input multi-task setups improve its explainability.