Using Topic Modelling to Personalise a Digital Self-compassion Training

Young adults that struggle with mental health issues experience barriers to seek help. With our online self-compassion training we try to overcome some of these barriers. To improve our training, we can personalise exercises based on topic modelling. Data from a pilot study is used to analyse and evaluate the algorithm. Overall, the algorithm has an accuracy of 54.1% for predicting the right topic. This accuracy increases to 80.4% when considering an empty prediction to be correct as well. Although this research also shows that our data makes the task of topic modelling difficult, it does prove to be a possibility to personalise the designed training.