| Summary: | Digital health platforms such as text messaging and mobile therapy are being increasingly embraced by patients as a valuable source of anxiety treatment. This mobile therapy-generated treatment comes in the form of text messages, and is normally known as short text and sparse. Since many real-world text-based data need semantic interpretation to reveal meaningful and relevant latent topics, research in Short Text Topic Modelling (STTM) was conducted. The current study examines the topics included in anxiety mobile therapy using STTM, particularly from the text messages sent by mental health professionals. Prior to the actual experiment, initial study was conducted using four topic modelling techniques with 28 text messages from anxiety therapy datasets and different hyperparameter settings. The performance evaluation includes classification accuracy, purity, normalized mutual information, and topic coherence. Based on the performance, Latent Feature Dirichlet Multinomial Mixture (LFDMM) with α = 0.1, β = 0.01, and K = 8 is found to be the most suitable hyperparameter setting for the anxiety text messages dataset and is used further in the actual experiment with 53 sample text. The findings from the actual experiment show that the anxiety text messages dataset comprises 8 interpretable topics that are classified under the domain of energy recharge, locus of control, mutual respect, activity scheduling, handling uncertainty, medium of communication, managing thoughts and health, and hope and readiness. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
|
|---|