Finding model through latent semantic approach to reveal the topic of discussion in discussion forum

摘要

There are lots of information and knowledge can be extracted from a discussion forum. Despite a discussion is opened by submitting a thread as the topic of discussion, however, the discussion may open out to different topics. This paper aims to present a model to find out a topic of discussion through latent semantic approach, named Topics Finding Model (TFM). The model proposes a complete step to reveal the topic of discussion from a thread in a discussion forum, consisting of the pre-processing text document, corpus classification and finding a topic. The model can be applied in various discussion forums and various languages with a few adjustments, such as stop-word removal list and stemming algorithm. The data were obtained from discussion forum in a learning management system. The data consist of 1050 posts divided into three different course subjects: information systems, management, and character building. The reason for using several course subjects is to observe consistency of the model. F-measure was used to measure the effectiveness of the model, and the results showed that the TFM was consistent and effective to reveal the topic of discussion, with a good precision. However, the recall can still be increased in a further study.