Mining App-Specific Opinions in Mobile App Reviews Using Lexico-Syntactic Patterns

EasyChair Preprint 15527

Abstract

Mobile app markets are increasingly crowded and competitive. User opinions in app reviews are important for app developers to improve users' experience and satisfaction with their apps. However, current approaches for automated analysis of mobile app reviews (e.g. SURF, AR-Miner) have two problems: firstly, developers do not have the ability to define their own topic of interest to query relevant information from app reviews, as these approaches focus more on the common and established aspects of the app (e.g. GUI, connection, payments, etc); secondly, while some of them use linguistic patterns to classify the intentions of each opinion, they suffered from low recall due to the messy nature of review text (e.g. SURF has 22.5% recall for feature request classification). To address these problems, in this paper, we introduce MARP as a semi-automatic tool for developers to search and classify their opinions of interest using a Lexico-Syntactic Pattern approach. A Lexico-Syntactic Pattern is a sequence of tokens including syntactic lexicons (e.g. 'have' or 'please') and content fragments (e.g. OBJECT or ACTION) that together represent the syntax pattern of an intention. Our experiments show that MARP can extract complaints and requests more accurately than the baseline approach and is rated with comparable usefulness by app developers.

Keyphrases: App Development, mobile app, phrase, syntactic pattern, topics and intentions, user reviews

@booklet{EasyChair:15527,
  author    = {Phong Vu and Tung Nguyen},
  title     = {Mining App-Specific Opinions in Mobile App Reviews Using Lexico-Syntactic Patterns},
  howpublished = {EasyChair Preprint 15527},
  year      = {EasyChair, 2024}}