Topic Modelling for Identification of Vaccine Reactions in Twitter

Detection of vaccine safety signals depends on various established reporting systems, where there is inevitably a lag between an adverse reaction to a vaccine and the reporting of it, and subsequent processing of reports. Therefore, it is desirable to try and detect safety signals earlier, ideally close to real-time. Extensive use of social media has provided a platform for sharing and seeking health-related information, and the immediacy of social media conversations mean that they are an ideal candidate for early detection of vaccine safety signals. The objective of this study is to evaluate topic models for identifying user posts on Twitter that most likely contain vaccine safety signals. This is an initial step in the overall research to determine if reliable vaccine safety signals can be detected in social media streams. The techniques used were focused on identifying the model design and number of topics that best revealed documents that contained vaccine safety signals, to assist with dimension reduction and subsequent labelling of the text data. The study compared Gensim LDA, MALLET, and jLDADMM DMM models to determine the most effective model for detecting vaccine safety signals, assisted by an evaluation process that used an adjusted F-Scoring technique over a labelled subset of the documents.