Topic modeling in software engineering research

Abstract Topic modeling using models such as Latent Dirichlet Allocation (LDA) is a text mining technique to extract human-readable semantic “topics” (i.e., word clusters) from corpus of textual documents. In software engineering, topic has been used analyze data in empirical studies (e.g., find out what developers talk about online), but also build new techniques support engineering tasks source code comprehension). needs be applied carefully depending on the type analyzed and parameters). Our study aims at describing how research with focus four aspects: (1) which have applied, (2) inputs for modeling, (3) was “prepared” pre-processed) (4) generated topics were named give them human-understandable meaning. We 111 papers ten highly-ranked venues (five journals five conferences) published between 2009 2020. found that LDA LDA-based are most frequent techniques, developer communication bug reports modelled most, pre-processing parameters vary quite bit often vaguely reported, manual naming (such deducting names based words topic) common.