 Schizophrenia is characterized by thought, language, and communication disorders. These impairments are often evident in patients' conversations with others. Assessments of thought disorder are critical for tracking the condition and detecting high-risk cases at an early stage. However, these assessments are time-consuming and expensive, requiring a trained clinician's expertise. To address this issue, we propose a machine-learning approach based on transformer-UBASE models to automate the assessment of thought disorder in schizophrenia patients. Our model extracts semantic, syntactic, and acoustic features from the interactions between occupational therapists slash psychiatric nurses and schizophrenia patients. It then uses these features to accurately predict the severity of the thought disorder. We conducted experiments on a data set of 35 patients and achieved promising results. This shows that our model can be used as a helpful tool for doctors when assessing schizophrenia patients. This article was authored by Yin Jia Huang, Yi Ting Lin, Xin Cheng Lu, and others. We are article.tv, links in the description below.