Utilizing BERT for Aspect-Based Sentiment Analysis via Constructing Auxiliary Sentence
2019-03-22NAACL 2019Code Available0· sign in to hype
Chi Sun, Luyao Huang, Xipeng Qiu
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/HSLCY/ABSA-BERT-pairOfficialIn paperpytorch★ 0
- github.com/recommeddit/labspytorch★ 1
- github.com/Mind23-2/MindCode-166none★ 0
- github.com/ywu94/Code-Notespytorch★ 0
- github.com/atharvajdhumal/Sentiment-Analysispytorch★ 0
- github.com/LorenzoAgnolucci/BERT_for_ABSApytorch★ 0
- github.com/anshulwadhawan/ABSApytorch★ 0
- github.com/mwbrulhardt/yelp-absapytorch★ 0
Abstract
Aspect-based sentiment analysis (ABSA), which aims to identify fine-grained opinion polarity towards a specific aspect, is a challenging subtask of sentiment analysis (SA). In this paper, we construct an auxiliary sentence from the aspect and convert ABSA to a sentence-pair classification task, such as question answering (QA) and natural language inference (NLI). We fine-tune the pre-trained model from BERT and achieve new state-of-the-art results on SentiHood and SemEval-2014 Task 4 datasets.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| SemEval 2014 Task 4 Subtask 4 | BERT-pair-QA-B | Accuracy (3-way) | 89.9 | — | Unverified |
| Sentihood | BERT-pair-QA-B | Aspect | 87.9 | — | Unverified |
| Sentihood | BERT-pair-QA-M | Aspect | 86.4 | — | Unverified |