Detecting health related discussions in everyday telephone conversations for studying medical events in the lives of older adults

We apply semi-supervised topic modeling techniques to detect health-related discussions in everyday telephone conversations, which has applications in large-scale epidemiological studies and for clinical interventions for older adults. The privacy requirements associ- ated with utilizing everyday telephone conversations preclude manual annotations; hence, we explore semi-supervised methods in this task. We adopt a semi-supervised version of Latent Dirichlet Allocation (LDA) to guide the learning process. Within this frame- work, we investigate a strategy to discard irrelevant words in the topic distribution and demonstrate that this strategy improves the average F-score on the in-domain task and an out-of-domain task (Fisher corpus). Our results show that the increase in the number of health-related conversations is statistically associated with actual medical events obtained through weekly self-reports.