Evaluating few-shot learning strategies for extracting feature-sentiment pairs from app reviews
Organization
Software Engineering Analytics
Abstract
Automatically analyzing user reviews to understand sentiments toward app functionality (i.e., app features) is essential for aligning development efforts with user expectations and needs. Recent advances in Large Language Models (LLMs), such as ChatGPT, have demonstrated impressive performance across several tasks without needing parameter updates, using zero or few labeled examples. Despite these advancements, the effectiveness of LLMs in performing feature-specific sentiment analysis of user reviews with 0-, 1-, and 5-shot settings [1] remains limited. This thesis aims to evaluate various few-shot learning strategies for extracting feature-sentiment pairs from app reviews to enhance the performance of small-scale LLMs.
[1] Shah, F. A., Sabir, A., & Sharma, R. (2024). A Fine-grained Sentiment Analysis of App Reviews using Large Language Models: An Evaluation Study. arXiv preprint arXiv:2409.07162.
[1] Shah, F. A., Sabir, A., & Sharma, R. (2024). A Fine-grained Sentiment Analysis of App Reviews using Large Language Models: An Evaluation Study. arXiv preprint arXiv:2409.07162.
Graduation Theses defence year
2024-2025
Supervisor
Faiz Ali Shah and Ahmed Sabir
Spoken language (s)
English
Requirements for candidates
Level
Bachelor, Masters
Application of contact
Name
Faiz Ali Shah
Phone
E-mail