Reinforcement Learning for Relation Classification From Noisy Data

TLDR

Existing relation classification methods based on distant supervision assume all sentences in a bag describe the relation, but bag‑level classification cannot map relations to individual sentences and suffers from noisy labeling. This paper proposes a novel sentence‑level relation classification model that learns from noisy data. The model comprises an instance selector that uses reinforcement learning to pick high‑quality sentences and a relation classifier that predicts relations at the sentence level and provides rewards, with both modules trained jointly. Experiments show the joint model effectively handles noisy data and achieves superior sentence‑level relation classification performance.

Abstract

Existing relation classification methods that rely on distant supervision assume that a bag of sentences mentioning an entity pair are all describing a relation for the entity pair. Such methods, performing classification at the bag level, cannot identify the mapping between a relation and a sentence, and largely suffers from the noisy labeling problem. In this paper, we propose a novel model for relation classification at the sentence level from noisy data. The model has two modules: an instance selector and a relation classifier. The instance selector chooses high-quality sentences with reinforcement learning and feeds the selected sentences into the relation classifier, and the relation classifier makes sentence-level prediction and provides rewards to the instance selector. The two modules are trained jointly to optimize the instance selection and relation classification processes.Experiment results show that our model can deal with the noise of data effectively and obtains better performance for relation classification at the sentence level.