Weakly Supervised Aspect Category Detection Method Based on Label Hierarchical Filtering Mechanism
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Aspect category detection is crucial for aspect-level sentiment analysis. Traditionally, supervised methodsrequire a large number of labeled samples to achieve effectiveness, whereas unsupervised methods rely heavilyon human judgment, which can compromise identification accuracy. To address these limitations, we proposea weakly supervised aspect category detection method that uses a hierarchical label filtering mechanism. Ourapproach begins by assigning preliminary pseudo-labels to comments based on topic similarity. Subsequently,unreliable pseudo-labeled samples are filtered out using semantic similarity. Finally, high-confidence trainingsamples are selected using both high and low thresholds. Using the hierarchical label filtering mechanismemployed in these three stages, we constructed a high-confidence training dataset, which was used to train atopic representation-enhanced classifier for aspect category detection. The proposed method, which wasevaluated on three publicly available datasets, outperformed existing baseline models, thereby reducing theneed for human intervention.
Abstract
Aspect category detection is crucial for aspect-level sentiment analysis. Traditionally, supervised methodsrequire a large number of labeled samples to achieve effectiveness, whereas unsupervised methods rely heavilyon human judgment, which can compromise identification accuracy. To address these limitations, we proposea weakly supervised aspect category detection method that uses a hierarchical label filtering mechanism. Ourapproach begins by assigning preliminary pseudo-labels to comments based on topic similarity. Subsequently,unreliable pseudo-labeled samples are filtered out using semantic similarity. Finally, high-confidence trainingsamples are selected using both high and low thresholds. Using the hierarchical label filtering mechanismemployed in these three stages, we constructed a high-confidence training dataset, which was used to train atopic representation-enhanced classifier for aspect category detection. The proposed method, which wasevaluated on three publicly available datasets, outperformed existing baseline models, thereby reducing theneed for human intervention.