Dąbrowski, Jacek; (2022) Mining app reviews to support software engineering. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text dabrowski_phd.pdf - Other Download (1MB) | Preview

Abstract

The thesis studies how mining app reviews can support software engineering. App reviews —short user reviews of an app in app stores— provide a potentially rich source of information to help software development teams maintain and evolve their products. Exploiting this information is however difficult due to the large number of reviews and the difficulty in extracting useful actionable information from short informal texts. A variety of app review mining techniques have been proposed to classify reviews and to extract information such as feature requests, bug descriptions, and user sentiments but the usefulness of these techniques in practice is still unknown. Research in this area has grown rapidly, resulting in a large number of scientific publications (at least 182 between 2010 and 2020) but nearly no independent evaluation and description of how diverse techniques fit together to support specific software engineering tasks have been performed so far. The thesis presents a series of contributions to address these limitations. We first report the findings of a systematic literature review in app review mining exposing the breadth and limitations of research in this area. Using findings from the literature review, we then present a reference model that relates features of app review mining tools to specific software engineering tasks supporting requirements engineering, software maintenance and evolution. We then present two additional contributions extending previous evaluations of app review mining techniques. We present a novel independent evaluation of opinion mining techniques using an annotated dataset created for our experiment. Our evaluation finds lower effectiveness than initially reported by the techniques authors. A final part of the thesis, evaluates approaches in searching for app reviews pertinent to a particular feature. The findings show a general purpose search technique is more effective than the state-of-the-art purpose-built app review mining techniques; and suggest their usefulness for requirements elicitation. Overall, the thesis contributes to improving the empirical evaluation of app review mining techniques and their application in software engineering practice. Researchers and developers of future app mining tools will benefit from the novel reference model, detailed experiments designs, and publicly available datasets presented in the thesis.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item