Detecting and tracking ongoing topics in psychotherapeutic conversations

One of the key aspects in a psychotherapeutic conversation is the understanding of topics dynamics driving the dialogue. This may provide insights on the therapeutic strategy adopted by the counselor for the specific patient, providing the opportunity of building up artificial intelligence (AI) based methods for recommending the most appropriate therapy for a new patient. In this paper, we propose a method able to detect and track topics in real-life psychotherapeutic conversations based on Partially Labeled Dirichlet Allocation. Topics detection helps in summarizing the semantic themes used during the therapeutic conversations, and in predicting a specific topic for each talk-turn. The conversation is modeled by means of a distribution of ongoing topics propagating through each talk sequence. Tracking of topics aims at exploring the dynamics of the conversation and at offering insights into the underlying conversation logic and strategy. We present an alternative way to look at face-to-face conversations in conjunction with a new approach that combines topic modeling and transitions matrices to elicit valuable knowledge.