Identifying and resolving conflict in mobile application features through contradictory feedback analysis

Area of Science:

• Computer Science
• Natural Language Processing
• Software Engineering

Background:

• Mobile applications generate vast amounts of user feedback.
• Identifying feature conflicts is essential for user satisfaction and app development.
• Existing methods struggle to effectively analyze and resolve conflicting feedback.

Purpose of the Study:

• To develop a novel methodology for identifying and resolving feature conflicts in mobile applications using NLP.
• To align mobile application development with user preferences and improve user experience.
• To provide actionable insights for prioritizing development requirements.

Main Methods:

• Utilized sentiment analysis (RoBERTa), topic modeling (Non-negative matrix factorization - NMF), and semantic similarity (Sentence-BERT).
• Detected conflicts by analyzing sentiment distributions within semantically similar feedback clusters.
• Incorporated antonym detection and negation handling for enhanced conflict detection.
• Employed majority voting, weighted ranking, and frequency analysis for conflict resolution.

Main Results:

• Successfully identified contradictory sentiments and latent topics representing app features.
• Clustered related feedback instances, enabling targeted conflict analysis.
• Demonstrated superior performance compared to baseline methods on large-scale datasets.
• Provided interpretable strategies for resolving feature conflicts.

Conclusions:

• The proposed NLP-driven methodology effectively identifies and resolves feature conflicts in mobile applications.
• This approach enhances mobile app development by aligning features with user preferences.
• The research offers a novel and effective strategy for improving user experience through feedback analysis.