VCounselor: A Psychological Intervention Chat Agent Based on a Knowledge-Enhanced Large Language Model

In the course of its work, the client's speeches are consolidated into a case, which is integrated into the structured input of the LLM together with dynamically changing additional material, and finally the LLM with structured fine-tuning is used to give a response to the client.

In order to achieve the effect of affective interaction with users, this paper uses Live2D technology to provide a movable avatar for VCounselor in the form of pseudo-holography. In addition, this paper is based on real-time speech recognition and emotional speech synthesis technology, which enables VCounselor to have more immediate and richer conversations with users. This would largely eliminate the uncanny valley effect and give the user the illusion of being in conversations with an agent with subjectivity.