FewRel: A Large-Scale Supervised Few-Shot Relation Classification Dataset with State-of-the-Art Evaluation

TL;DR

FewRel is a large-scale, challenging dataset for few-shot relation classification derived from Wikipedia, revealing that current models still lag behind human performance and highlighting the need for further research.

Contribution

The paper introduces FewRel, a new large-scale dataset for few-shot relation classification, and evaluates state-of-the-art methods, exposing their limitations and suggesting future research directions.

Findings

Current models perform significantly worse than humans.

Few-shot relation classification remains an open challenge.

Different reasoning skills are required to solve the task.

Abstract

We present a Few-Shot Relation Classification Dataset (FewRel), consisting of 70, 000 sentences on 100 relations derived from Wikipedia and annotated by crowdworkers. The relation of each sentence is first recognized by distant supervision methods, and then filtered by crowdworkers. We adapt the most recent state-of-the-art few-shot learning methods for relation classification and conduct a thorough evaluation of these methods. Empirical results show that even the most competitive few-shot learning models struggle on this task, especially as compared with humans. We also show that a range of different reasoning skills are needed to solve our task. These results indicate that few-shot relation classification remains an open problem and still requires further research. Our detailed analysis points multiple directions for future research. All details and resources about the dataset and baselines are released on http://zhuhao.me/fewrel.