STAGE: Span Tagging and Greedy Inference Scheme for Aspect Sentiment Triplet Extraction

• DOI: 10.1609/aaai.v37i11.26547
• Publication Date: 2023-06-27T18:09:04Z
• Authors: Shuo Liang; Wei Wei; Xian-Ling Mao; Yuanyuan Fu; Rui Fang; Dangyang Chen
• Keywords: Sentiment Analysis; Margin (machine learning)

Aspect Sentiment Triplet Extraction (ASTE) has become an emerging task in sentiment analysis research, aiming to extract triplets of the aspect term, its corresponding opinion and associated polarity from a given sentence. Recently, many neural networks based models with different tagging schemes have been proposed, but almost all them their limitations: heavily relying on 1) prior assumption that each word is only single role (e.g., or etc. ) 2) word-level interactions treating opinion/aspect as set independent words. Hence, they perform poorly complex ASTE task, such multiple roles aspect/opinion term we propose novel approach, Span TAgging Greedy infErence (STAGE), span-level, where span may consist words play simultaneously. To this end, paper formulates multi-class classification problem. Specifically, STAGE generates more accurate triplet extractions via exploring span-level information constraints, which consists two components, namely, scheme greedy inference strategy. The former tag possible candidate spans newly-defined set. latter retrieves maximum length snippet output triplets. Furthermore, simple effective model STAGE, outperforms state-of-the-arts by large margin four widely-used datasets. Moreover, our can be easily generalized other pair/triplet extraction tasks, also demonstrates superiority proposed STAGE.