Preparing Data from Psychotherapy for Natural Language Processing

Preparing Data from Psychotherapy for Natural Language Processing Margot Mieskes, Andreas Stiegelmayr Hochschule Darmstadt Darmstadt, Germany {firstname.lastname}@h-da.de Abstract Mental health and well-being are growing issues in western civilizations. But at the same time, psychotherapy and further edu- cation in psychotherapy is a highly demanding occupation, resulting in a severe gap in patient-centered care. The question which arises from recent developments in natural language processing (NLP) and speech recognition is, how these technologies could be employed to support the therapists in their work and allow for a better treatment of patients. Most research in NLP focuses on analysing the language of patients with various psychological conditions, but only few examples exist that analyse the therapists behavior and the interaction between therapist and patient. We present ongoing work in collecting, preparing and analysing data from psychotherapy sessions together with expert annotations on various qualitative dimensions of these sessions, such as feed- back and cooperation. Our aim is to use this data in a classification task, which gives insight into what qualifies for good feedback or cooperation in therapy sessions and employ this information to support psychotherapists in improving the quality of the care they offer. Keywords: psychotherapy, speech resources, German, data preparation, guidelines, data preprocessing 1. Introduction Mental health care is a very demanding occupation, where quality assurance and improvement is very time- and re- source intensive. Improving mental health care through psychotherapy involves two aspects: First, the point of view of the therapist, who is required to constantly attend further education measures, to ensure the quality of their work remains high and by improving it through supervi- sion. Second, the patients’ point of view, who has a right for high quality care, especially for conditions, where the treatment greatly benefits from psychotherapy. But due to a lack of therapists, these patients are still treated with drugs, which are less effective. As there is a lack in fur- ther education options for therapists, the number of qual- ified therapists is hardly increasing, resulting in a vicious circle. This is especially apparent in the treatment of pa- tients with schizophrenic conditions. To improve the sit- uation, psychological therapists are collecting data from therapy sessions, which are then judged by experts from their field. Their guidelines define 14 quality dimentions, such as Feedback or Cooperation, rated on a seven-point Likert scale. This work is currently carried out manually and is very time-consuming and resource extensive. Previ- ous work on applying Natural Language Processing (NLP) shows that NLP can be used to extract certain aspects in human-human communication. In our work, we build on this previous work to support the analysis of the therapy session. Before this, the data has to be processed, which requires methods and tools, we partially develop ourselves. Our work describes efforts in creating a pipeline for prepro- cessing therapy recordings to use NLP and machine learn- ing to support psychotherapists in analysing their work. This preprocessing itself poses several challenges: As pa- tients are quite sensitive to their surroundings, the …

Content truncated to 500 words