🤖 AI Summary
Conventional sentiment analysis treats aspect category detection, aspect-level sentiment classification, and rating prediction as three isolated tasks, ignoring their intrinsic interdependencies and relying on costly fine-grained annotations.
Method: This paper proposes a unified sentiment analysis paradigm leveraging only coarse-grained star ratings for weakly supervised learning. Its core innovation is the Distant Supervision Pyramid Network (DSPN), which jointly optimizes all three tasks via multi-level (word-, aspect-, and document-level) feature modeling. A pyramid architecture enables multi-granularity sentiment representation while ensuring output interpretability. The model supports cross-lingual joint training for English and Chinese.
Contribution/Results: Evaluated on multi-aspect review datasets, our approach achieves performance comparable to state-of-the-art fully supervised models—despite requiring significantly lower annotation effort—thereby substantially improving efficiency, generalizability, and interpretability.
📝 Abstract
Sentiment analysis is integral to understanding the voice of the customer and informing businesses’ strategic decisions. Conventional sentiment analysis involves three separate tasks: aspect-category detection, aspect-category sentiment analysis, and rating prediction. However, independently tackling these tasks can overlook their interdependencies and often requires expensive, fine-grained annotations. This paper introduces unified sentiment analysis, a novel learning paradigm that integrates the three aforementioned tasks into a coherent framework. To achieve this, we propose the Distantly Supervised Pyramid Network (DSPN), which employs a pyramid structure to capture sentiment at word, aspect, and document levels in a hierarchical manner. Evaluations on multi-aspect review datasets in English and Chinese show that DSPN, using only star rating labels for supervision, demonstrates significant efficiency advantages while performing comparably well to a variety of benchmark models. Additionally, DSPN’s pyramid structure enables the interpretability of its outputs. Our findings validate DSPN’s effectiveness and efficiency, establishing a robust, resource-efficient, unified framework for sentiment analysis.
