English

Fine-tuning Pre-Trained Transformer Language Models to Distantly Supervised Relation Extraction

Computation and Language 2019-06-21 v1

Abstract

Distantly supervised relation extraction is widely used to extract relational facts from text, but suffers from noisy labels. Current relation extraction methods try to alleviate the noise by multi-instance learning and by providing supporting linguistic and contextual information to more efficiently guide the relation classification. While achieving state-of-the-art results, we observed these models to be biased towards recognizing a limited set of relations with high precision, while ignoring those in the long tail. To address this gap, we utilize a pre-trained language model, the OpenAI Generative Pre-trained Transformer (GPT) [Radford et al., 2018]. The GPT and similar models have been shown to capture semantic and syntactic features, and also a notable amount of "common-sense" knowledge, which we hypothesize are important features for recognizing a more diverse set of relations. By extending the GPT to the distantly supervised setting, and fine-tuning it on the NYT10 dataset, we show that it predicts a larger set of distinct relation types with high confidence. Manual and automated evaluation of our model shows that it achieves a state-of-the-art AUC score of 0.422 on the NYT10 dataset, and performs especially well at higher recall levels.

Keywords

Cite

@article{arxiv.1906.08646,
  title  = {Fine-tuning Pre-Trained Transformer Language Models to Distantly Supervised Relation Extraction},
  author = {Christoph Alt and Marc Hübner and Leonhard Hennig},
  journal= {arXiv preprint arXiv:1906.08646},
  year   = {2019}
}

Comments

To appear in Proceedings of ACL 2019 (11 pages). arXiv admin note: text overlap with arXiv:1906.03088

R2 v1 2026-06-23T09:59:02.773Z