Label Verbalization and Entailment for Effective Zero- and Few-Shot Relation Extraction
This work addresses the data annotation bottleneck for relation extraction, offering a practical solution with significant gains in low-data regimes, though it is incremental as it builds on existing entailment models.
The authors tackled the problem of costly labeled data for relation extraction by reformulating it as an entailment task using hand-made verbalizations, achieving 63% F1 zero-shot and 69% F1 with 16 examples per relation, which is 17% better than the best supervised system under the same conditions.
Relation extraction systems require large amounts of labeled examples which are costly to annotate. In this work we reformulate relation extraction as an entailment task, with simple, hand-made, verbalizations of relations produced in less than 15 min per relation. The system relies on a pretrained textual entailment engine which is run as-is (no training examples, zero-shot) or further fine-tuned on labeled examples (few-shot or fully trained). In our experiments on TACRED we attain 63% F1 zero-shot, 69% with 16 examples per relation (17% points better than the best supervised system on the same conditions), and only 4 points short to the state-of-the-art (which uses 20 times more training data). We also show that the performance can be improved significantly with larger entailment models, up to 12 points in zero-shot, allowing to report the best results to date on TACRED when fully trained. The analysis shows that our few-shot systems are specially effective when discriminating between relations, and that the performance difference in low data regimes comes mainly from identifying no-relation cases.