Aspect Sentiment Triplet Extraction on Lap SemEval 2014 (test)
68.33F1 ScoreEMC-GCN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| EMC-GCNParadigm=Table tagging2026.05 | 68.33 | 65.62 | 71.3 | |
| MvP + FiVeD-unifiedVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 64.38 | 66.42 | 62.51 | |
| MvP + FiVeDVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 63.86 | 68.09 | 60.14 | |
| MvP + FiVeDVerification Model=Qwen, Paradigm=Generative2026.05 | 63.73 | 68.58 | 59.52 | |
| D2E2SParadigm=Span-based2026.05 | 63.65 | 67.38 | 60.31 | |
| MiniConGTSParadigm=Table tagging2026.05 | 63.61 | 66.82 | 60.68 | |
| MvP + FiVeD-unifiedVerification Model=Qwen, Paradigm=Generative2026.05 | 63.6 | 66.76 | 60.78 | |
| TAGSParadigm=Sequence tagging2026.05 | 63.26 | 64.69 | 61.89 | |
| MvPParadigm=Generative2026.05 | 62.73 | 65.86 | 59.89 | |
| Paraphrase + FiVeDVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 62.21 | 66.87 | 58.16 | |
| LEGOParadigm=Generative2026.05 | 62.2 | — | — | |
| SimSTARParadigm=Table tagging2026.05 | 62.07 | 66.46 | 58.23 | |
| Paraphrase + FiVeD-unifiedVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 62.04 | 67.33 | 57.53 | |
| Paraphrase + FiVeDVerification Model=Qwen, Paradigm=Generative2026.05 | 61.98 | 65.93 | 58.53 | |
| Paraphrase + FiVeD-unifiedVerification Model=Qwen, Paradigm=Generative2026.05 | 61.75 | 66.47 | 57.68 | |
| PARAPHRASEType=Sequence-to-sequence, Modeling=Natural Language Labels2021.10 | 61.13 | — | — | |
| GAS + FiVeD-unifiedVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 60.99 | 58.33 | 64.16 | |
| GAS + FiVeDVerification Model=Qwen, Paradigm=Generative2026.05 | 60.94 | 59.36 | 62.73 | |
| GAS + FiVeD-unifiedVerification Model=Qwen, Paradigm=Generative2026.05 | 60.88 | 58.82 | 63.35 | |
| GAS + FiVeDVerification Model=DeepSeek-Reasoner, Paradigm=Generative2026.05 | 60.86 | 57.99 | 64.09 | |
| Triple-MRCParadigm=MRC-based2026.05 | 60.72 | — | — | |
| ParaphraseParadigm=Generative2026.05 | 60.41 | 64.14 | 56.06 | |
| COM-MRCParadigm=MRC-based2026.05 | 60.17 | 62.35 | 58.16 | |
| Span-ASTEEncoder=BERT2021.07 | 59.38 | 63.44 | 55.84 | |
| Span-ASTEParadigm=Span-based2026.05 | 59.38 | 63.44 | 55.84 | |
| Span-BARTParadigm=Sequence tagging2026.05 | 58.69 | 61.41 | 56.19 | |
| Two-StageType=Two-stage2021.10 | 58.58 | — | — | |
| GASType=Sequence-to-sequence2021.10 | 58.19 | — | — | |
| GASParadigm=Generative2026.05 | 58.03 | 59.89 | 56.43 | |
| PBF-M12021.03 | 56.3 | 58.8 | 54.1 | |
| PBF2021.03 | 55.8 | 56.6 | 55.1 | |
| PBF -w/o A2021.03 | 55.8 | 55.9 | 55.7 | |
| PBF -w/o P2021.03 | 55.7 | 56.6 | 54.8 | |
| Dual-MRCParadigm=MRC-based2026.05 | 55.58 | 57.39 | 53.88 | |
| GTS+BERTType=End-to-end, Backbone=BERT2021.10 | 55.21 | — | — | |
| PBF-M32021.03 | 54.9 | 56.8 | 53.3 | |
| GTSEncoder=BERT2021.07 | 54.36 | 57.82 | 51.32 | |
| GTS-BERT2021.03 | 53.5 | 59.4 | 48.6 | |
| Span-ASTEEncoder=BiLSTM2021.07 | 51.8 | 59.85 | 45.67 | |
| JETEncoder=BERT2021.07 | 51.04 | 55.39 | 47.33 | |
| Jet+BERTType=End-to-end, Backbone=BERT2021.10 | 51.04 | — | — | |
| PBF-M22021.03 | 49.4 | 51.8 | 47.3 | |
| PBF -w/o AP2021.03 | 46.7 | 45.1 | 48.8 | |
| GTSEncoder=BiLSTM2021.07 | 46.31 | 53.35 | 40.99 | |
| JET-bert-o2021.03 | 46 | 49.7 | 42.8 | |
| JET-bert-t2021.03 | 45.5 | 50.2 | 41.7 | |
| Zhang et al.Encoder=BiLSTM2021.07 | 44.78 | 49.62 | 41.07 | |
| GTS-CNN2021.03 | 44.6 | 55.3 | 37.4 | |
| GPT-3.5-turboEvaluation Prompting Strategy=few-shot, Paradigm=Generative2026.05 | 44.35 | 39.79 | 50.09 | |
| Peng et al.Encoder=BiLSTM2021.07 | 42.87 | 37.38 | 50.38 | |
| P-pipelineType=Pipeline-based2021.10 | 42.87 | — | — | |
| Peng-two-stageParadigm=Sequence tagging2026.05 | 42.87 | 37.38 | 50.38 | |
| GTS-BiLSTM2021.03 | 42.8 | 58 | 33.9 | |
| GPT-4oEvaluation Prompting Strategy=few-shot, Paradigm=Generative2026.05 | 42.8 | 38.23 | 48.61 | |
| OTE-MTL2021.03 | 42.7 | 51.3 | 36.8 | |
| Li-unified-REncoder=BiLSTM2021.07 | 42.34 | 40.56 | 44.28 | |
| Li-unified-RType=Pipeline-based2021.10 | 42.34 | — | — | |
| JETo2021.03 | 41.9 | 48.8 | 36.7 | |
| JETEncoder=BiLSTM2021.07 | 41.35 | 53.03 | 33.89 | |
| JETt2021.03 | 38.6 | 41 | 36.5 | |
| GPT-3.5-turboEvaluation Prompting Strategy=CoT + few-shot, Paradigm=Generative2026.05 | 37.57 | 33.78 | 42.33 | |
| GPT-3.5-turboEvaluation Prompting Strategy=CoT, Paradigm=Generative2026.05 | 34.71 | 30.48 | 40.3 | |
| GPT-3.5-turboEvaluation Prompting Strategy=zero-shot, Paradigm=Generative2026.05 | 34.69 | 30.04 | 41.04 | |
| GPT-4oEvaluation Prompting Strategy=CoT + few-shot, Paradigm=Generative2026.05 | 34.4 | 29.71 | 40.85 | |
| CMLA+Encoder=BiLSTM2021.07 | 33.16 | 30.09 | 36.92 | |
| CMLA+Type=Pipeline-based2021.10 | 33.16 | — | — | |
| GPT-4oEvaluation Prompting Strategy=CoT, Paradigm=Generative2026.05 | 31.46 | 26.98 | 37.71 | |
| RINANTE+Encoder=BiLSTM2021.07 | 20.07 | 21.71 | 18.66 | |
| GPT-4oEvaluation Prompting Strategy=zero-shot, Paradigm=Generative2026.05 | 19.9 | 17.81 | 22.55 |