The therapeutic working alliance is a critical factor in predicting the success of psychotherapy treatment. Traditionally, working alliance assessment relies on questionnaires completed by both therapists and patients. In this paper, we present COMPASS, a novel framework to directly infer the therapeutic working alliance from the natural language used in psychotherapy sessions. Our approach utilizes advanced large language models to analyze transcripts of psychotherapy sessions and compare them with distributed representations of statements in the working alliance inventory. Analyzing a dataset of over 950 sessions covering diverse psychiatric conditions, we demonstrate the effectiveness of our method in microscopically mapping patient-therapist alignment trajectories and providing interpretability for clinical psychiatry and in identifying emerging patterns related to the condition being treated. By employing various neural topic modeling techniques in combination with generative language prompting, we analyze the topical characteristics of different psychiatric conditions and incorporate temporal modeling to capture the evolution of topics at a turn-level resolution. This combined framework enhances the understanding of therapeutic interactions, enabling timely feedback for therapists regarding conversation quality and providing interpretable insights to improve the effectiveness of psychotherapy.

Long been a clinical quantity estimated by the patients' (WAT), a transformer-based classification model to classify and therapists' self-evaluative reports, we believe that the the psychotherapy sessions into different psychiatric working alliance can be better characterized using natural conditions. Our methods consists of a psychological state language processing technique directly in the dialogue transcribed encoder that quantifies the degree of patient-therapist alliance in each therapy session. In this work, we propose the by projecting each turn in a therapeutic session onto Working Alliance Transformer (WAT), a Transformer-based the representation of clinically established working alliance classification model that has a psychological state encoder inventories, using language modeling to encode both turns which infers the working alliance scores by projecting the and inventories, which was originally proposed in [4] as an embedding of the dialogues turns onto the embedding space analytical tool. This allows us not only to quantify the overall of the clinical inventory for working alliance. We evaluate our degree of alliance but also to identify granular patterns its dynamics method in a real-world dataset with over 950 therapy sessions over shorter and longer time scales.