TASCI: transformers for aspect-based sentiment analysis with contextual intent integration.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Computational Linguistics

Background:

• Traditional sentiment analysis often fails to capture the relationship between user intent and sentiment towards specific aspects.
• Existing models may overlook the contextual nuances crucial for accurate opinion mining.

Purpose of the Study:

• To introduce a novel Transformer-Based Aspect-Level Sentiment Classification with Intent (TASCI) model.
• To enhance aspect-level sentiment classification by integrating intent analysis.
• To improve the accuracy and contextual understanding of sentiment analysis.

Main Methods:

• Aspect extraction using a self-attention mechanism.
• Intent inference from preceding sentences using a Transformer-based model.
• Integration of aspect and intent analysis for a unified sentiment classification framework.

Main Results:

• Achieved state-of-the-art results on benchmark datasets (Restaurant, Laptop, Twitter).
• Demonstrated high accuracy: 89.10% (Restaurant), 84.81% (Laptop), 79.08% (Twitter).
• Achieved high macro-F1 scores: 83.38% (Restaurant), 78.63% (Laptop), 77.27% (Twitter).

Conclusions:

• Incorporating intent analysis significantly improves aspect-level sentiment classification.
• The TASCI model provides a more accurate reflection of user opinions by contextualizing sentiment.
• The findings establish a new standard for complex sentiment expression analysis across diverse domains.