RE-SWOT: From User Feedback to Requirements via Competitor Analysis

Abstract

Context & Motivation] App store reviews are a rich source for analysts to elicit requirements from user feedback, for they describe bugs to be fixed, requested features, and possible improvements. Product development teams need new techniques that help them make real-time decisions based on user feedback. [Question/Problem] Researchers have proposed natural language processing (NLP) techniques for extracting and organizing requirements-relevant knowledge from the reviews for one specific app. However, no attention has been paid to studying whether and how requirements can be identified from competing products. [Principal ideas/results] We propose RE-SWOT, a tool-supported method for eliciting requirements from app store reviews through competitor analysis. RE-SWOT combines NLP algorithms with information visualization techniques. We evaluate the usefulness of RE-SWOT with expert product managers from three mobile app companies.[Contribution] Our preliminary results show that competitor analysis is a promising path for research that has direct impact on the requirements engineering practice in modern app development companies.

Mining User Opinions to Support Requirement Engineering: An Empirical Study

“…Our study indicates that feature extraction techniques are not yet effective enough to be used in practice [9,21] and that have lower precision and recall than reported in their initial studies. Our study also indicates that feature-specific sentiment analysis techniques have limited precision and recall, particularly for negative sentiments.…”

“…In our study, we selected three approaches: GuMa [12], SAFE [13] and ReUS [10]. We selected GuMa and SAFE as they are state-of-the-art approaches widely known in RE community [9,17,22]. We opted for ReUS [10] as the approach achieves a competitive performance in the context of opinion mining and sentiment analysis research [10,16].…”

Data-Driven Requirements Elicitation: A Systematic Literature Review

“…Online reviews are the most frequently used type of dynamic data for eliciting requirements (53%), followed by micro-blogs (18%) and online discussions/forums (12%), software repositories Online reviews • Online reviews included app reviews, reviews compiled by experts, and online user reviews. Among the studies which used online reviews, a majority of the studies used app reviews as the sources of potential requirements (75%) [35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54]. Of them, 14 used app reviews from multiple distribution platforms such as Apple AppStore and Google Play to increase the level of generalizability, while eleven used those from a single distribution platform, and one did not specify the number of app distribution platforms • Of the studies which used online reviews, 17% (n = 6) extracted user reviews of software and video games [55], IoT products [56], compact cameras [57], internet security [58], Jira and Trello [59], and Jingdong.…”

Assisted authoring of model-based systems engineering documents

“…In Requirements Engineering, the analysis of user reviews is a rich and growing source of information [3]. In [4], Dalpiaz et al…”

“…Our implementation uses spaCy 3 , an open-source NLP library with a python API, and a pretrained model for the English language trained by written texts from the web including blogs, news and comments 4 . The NLP pipeline we set up for this experiment consists of three stages: (1) Tokenization, which uses a pretrained English model to turn a series of characters into a structured series of words/phrases; (2) Part-of-Speech (POS) tagging, which assigns each token a part-of-speech label such as "noun", "verb", "adjective", and so on; and (3) noun-chunk merging, where the pretrained English model forms noun-phrases by merging adjacent tokens that likely refer to an entity, e.g., "The Golden Gate Bridge".…”