Short Text Topic Modeling Techniques, Applications, and Performance: A Survey

Analyzing short texts infers discriminative and coherent latent topics that is a critical fundamental task since many real-world applications require semantic understanding of texts. Traditional long text topic modeling algorithms (e.g., PLSA LDA) based on word co-occurrences cannot solve this problem very well only limited co-occurrence information available in Therefore, has already attracted much attention from the machine learning research community recent years, which aims at overcoming sparseness In survey, we conduct comprehensive review various techniques proposed literature. We present three categories methods Dirichlet multinomial mixture, global co-occurrences, self-aggregation, with example representative approaches each category analysis their performance tasks. develop first open-source library, called STTM, for use Java integrates all surveyed within unified interface, benchmark datasets, to facilitate expansion new field. Finally, evaluate these state-of-the-art datasets compare against one another versus algorithm.