Aspect-Based Sentiment Analysis on Restaurant14 original (test)
92.68AccuracyMT-ISA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MT-ISAModel=Flan-T5 XXL (11B), Strategy=Output Strategy (O)2024.12 | 92.68 | 88.96 | |
| MT-ISAModel=Flan-T5 XXL (11B), Strategy=Input Strategy (I)2024.12 | 91.34 | 86.71 | |
| Direct fine-tuneModel=Flan-T5 XXL (11B), Protocol=Prompt-based fine-tuning2024.12 | 89.29 | 83.68 | |
| BERT Asp + SCAPT†Model=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 89.11 | 83.79 | |
| THORModel=Flan-T5 XXL (11B), Protocol=Prompt-based fine-tuning2024.12 | 88.57 | 82.93 | |
| ABSA-ESA*Model=T5-Base (220M), Protocol=Fine-tuned2024.12 | 88.29 | 81.74 | |
| MT-ISAModel=Flan-T5 Base (250M), Strategy=Input Strategy (I)2024.12 | 88.21 | 82.45 | |
| THORModel=Flan-T5 Base (250M), Protocol=Prompt-based fine-tuning2024.12 | 87.68 | 81.1 | |
| BERT Asp + CEPT†Model=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 87.5 | 82.07 | |
| BERT-ADA+Model=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 87.14 | 80.05 | |
| ISAIVModel=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 87.05 | 81.4 | |
| C³DA+Model=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 86.93 | 81.23 | |
| Direct fine-tuneModel=Flan-T5 Base (250M), Protocol=Prompt-based fine-tuning2024.12 | 86.88 | 79.78 | |
| RGATModel=BERT-Large (340M), Protocol=Fine-tuned2024.12 | 86.87 | 79.99 | |
| RGATModel=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 86.6 | 81.35 | |
| MT-ISAModel=Flan-T5 Base (250M), Strategy=Output Strategy (O)2024.12 | 85.89 | 77.85 | |
| GPT-4o-miniProtocol=Zero-shot2024.12 | 84.82 | 72.52 | |
| BERT-SPC+Model=BERT-Base (110M), Protocol=Fine-tuned2024.12 | 83.57 | 77.16 | |
| GPT-4o-mini + THORProtocol=Zero-shot2024.12 | 81.25 | 62.12 | |
| InstructABSAModel=TK-INSTRUCT (200M), Protocol=Prompt-based fine-tuning2024.12 | 80.89 | 38.26 |