Dynamic Order Template Prediction for Generative Aspect-Based Sentiment Analysis

2024-06-17

arXiv.org

AI Summary

Existing static template-based approaches for aspect-based sentiment analysis (ABSA) struggle to capture interdependencies among sentiment tuple elements (e.g., aspect, opinion, sentiment polarity), while multi-view prompting methods suffer from inefficient inference and poor out-of-distribution robustness. To address these limitations, this paper proposes a dynamic sequential template mechanism that adaptively selects the optimal template sequence per instance based on prediction entropy, thereby jointly optimizing both diversity and relevance. Our approach introduces, for the first time, an entropy-driven dynamic view generation strategy, integrating generative modeling, dynamic template learning, and multi-task joint decoding. Extensive experiments on the ASQP and ACOS benchmarks demonstrate significant F1-score improvements, substantial reduction in inference latency, and—crucially—the first simultaneous achievement of state-of-the-art accuracy and efficiency in ABSA.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Multi-instance/Multi-view Learning
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Aspect-based sentiment analysis (ABSA) assesses sentiments towards specific aspects within texts, resulting in detailed sentiment tuples. Previous ABSA models often use static templates to predict all of the elements in the tuples, and these models often fail to accurately capture dependencies between elements. Multi-view prompting method improves the performance of ABSA by predicting tuples with various templates and then ensembling the results. However, this method suffers from inefficiencies and out-of-distribution errors. In this paper, we propose a Dynamic Order Template (DOT) method for ABSA, which dynamically generates necessary views for each instance based on instance-level entropy. Ensuring the diverse and relevant view generation, our proposed method improves F1-scores on ASQP and ACOS datasets while significantly reducing inference time.

Problem

• Improves aspect-based sentiment analysis accuracy
• Reduces inefficiencies in multi-view prompting
• Dynamically generates relevant views per instance

Innovation

• Dynamic Order Template for ABSA
• Instance-level entropy for view generation
• Improves F1-scores and reduces time