Using Chief Complaint for Speech Analytics and Text Mining to Assess Cognitive functions in Patients with Depressive Disorders and Bipolar Disorders

Depression and bipolar disorder are two common and important mood disorders, and patients often have obvious functional impairment. Recent studies have shown that even when the mood is stable, the cognitive function of patients is worse than that of the healthy group, and cognitive impairment may be the cause of the deterioration of patients' occupational function and interpersonal relationship. However, in actual clinical situations, the cognitive function of patients is rarely evaluated regularly. One of the main reasons is that the evaluation of cognitive function usually requires additional arrangements for cognitive function testing, which is often time-consuming and labor-intensive. The cognitive function status of patients with depressive disorder and bipolar disorder is often an overlooked issue.In the past, there have been studies using text mining, audio analysis, facial emotion recognition and other methods to identify potential patients with depression, and hope to find patients with depression as early as possible for early intervention. However, through literature review, we can find that there is no research using text mining, audio analysis and other techniques to analyze the cognitive function status of patients with depressive disorder or bipolar disorder.Therefore, this study attempts to combine the characteristics and methods of text and speech data to construct a predictive model of the degree of cognitive impairment in patients with depressive disorder or bipolar disorder.The study is expected to allow patients to dictate their recent mood states and record them. At the same time, the cognitive function assessment will be done to assess the degree of cognitive impairment in the case. The spoken audio files will be used for speech exploration and text exploration to construct the predictive model for cognition impairment of patients with depressive disorder or bipolar disorder. In addition, it also analyzes the voice content data of patients with depressive disorder and bipolar disorder, trying to establish a predictive model to distinguish between patients with depressive disorder and bipolar depression.