View Proposal

---

Proposer

Timothy Yap

Title

Refining Aspect-Based Sentiment Analysis Through Subjectivity-Aware Multitask Learning

Goal

To develop and evaluate a multitask learning framework that jointly performs Subjectivity Analysis and Aspect-Based Sentiment Analysis (ABSA), in order to improve the accuracy and granularity of sentiment interpretation in customer feedback for a specific product or service domain.

Description

Aspect-Based Sentiment Analysis (ABSA) is a fine-grained sentiment analysis technique that identifies specific aspects of a product, service, or entity within text and determines the sentiment expressed toward each aspect. Unlike traditional sentiment analysis, which focuses on overall sentiment, ABSA provides more detailed and actionable insights. Subjectivity Analysis focuses on distinguishing subjective content, which expresses personal opinions, emotions, or evaluations, from objective content that primarily conveys factual information. Since sentiments are typically expressed within subjective statements, subjectivity information can provide valuable contextual signals for sentiment interpretation. This project aims to design, implement, and evaluate a multitask learning framework that jointly learns Subjectivity Analysis and Aspect-Based Sentiment Analysis. Rather than treating subjectivity as a separate preprocessing step in a pipeline, both tasks will be learned simultaneously through shared representations within a neural network model. The project will investigate whether incorporating subjectivity awareness as an auxiliary task can improve aspect-level sentiment prediction performance. The resulting system will be evaluated on customer feedback from a specific domain, with the goal of generating more accurate and nuanced insights into user opinions and concerns.

Resources

Python, Deep Neural Networks, Transformer-based Language Models (e.g., BERT, RoBERTa)

Background

Machine Learning, Deep Learning, Natural Language Processing (NLP), Python Programming

Url

Difficulty Level

Moderate

Ethical Approval

Full

Number Of Students

1

Supervisor

Timothy Yap

Keywords

aspect-based sentiment analysis, subjectivity analysis, multitask learning, deep learning, neural networks, natural language processing, sentiment analysis

Degrees

Bachelor of Science in Computing Science