Distilroberta2gnn: a new hybrid deep learning approach for aspect-based sentiment analysis.

Area of Science:

• Natural Language Processing (NLP)
• Machine Learning
• Artificial Intelligence

Background:

• Aspect-Based Sentiment Analysis (ABSA) is vital for understanding opinions on specific topics within text.
• Current ABSA methods struggle with subtle language nuances and a lack of high-quality, domain-specific datasets.
• Existing models often lack the adaptability required for diverse linguistic contexts.

Purpose of the Study:

• To introduce and evaluate the Distil-RoBERTa2GNN model, a novel hybrid approach for ABSA.
• To address challenges in interpreting nuanced language and the scarcity of annotated datasets in ABSA.
• To establish a new performance benchmark for ABSA research.

Main Methods:

• Developed a hybrid model combining DistilRoBERTa for feature extraction and Graph Neural Networks (GNNs) for classification.
• Implemented a comprehensive four-phase data preprocessing strategy to enhance training data quality.
• Evaluated the model on four benchmark datasets: Rest14, Rest15, Rest16-EN, and Rest16-ESP.

Main Results:

• Achieved strong F1 scores across datasets: Rest14 (77.98%), Rest15 (76.86%), Rest16-EN (84.96%), and Rest16-ESP (74.87%).
• Demonstrated competitive performance compared to various baseline models in ABSA.
• Highlighted the model's effectiveness in handling domain-specific sentiment analysis tasks.

Conclusions:

• The Distil-RoBERTa2GNN model offers a robust solution for aspect-based sentiment analysis.
• The hybrid approach effectively leverages pre-trained language models and GNNs for improved sentiment classification.
• This research advances ABSA by enhancing model adaptability and performance on diverse datasets.