Weakly-supervised Relation Extraction by Pattern-enhanced Embedding Learning

Extracting relations from text corpora is an important task in text mining. It becomes particularly challenging when focusing on weakly-supervised relation extraction, that is, utilizing a few relation instances (i.e., a pair of entities and their relation) as seeds to extract more instances from corpora. Existing distributional approaches leverage the corpus-level co-occurrence statistics of entities to predict their relations, and require large number of labeled instances to learn e‚ective relation classi€ers. Alternatively, paŠern-based approaches perform bootstrapping or apply neural networks to model the local contexts, but still rely on large number of labeled instances to build reliable models. In this paper, we study integrating the distributional and paŠern-based methods in a weakly-supervised seŠing, such that the two types of methods can provide complementary supervision for each other to build an e‚ective, uni€ed model. We propose a novel co-training framework with a distributional module and a paŠern module. During training, the distributional module helps the paŠern module discriminate between the informative paŠerns and other paŠerns, and the paŠern module generates some highly-con€dent instances to improve the distributional module. Œe whole framework can be e‚ectively optimized by iterating between improving the paŠern module and updating the distributional module. We conduct experiments on two tasks: knowledge base completion with text corpora and corpus-level relation extraction. Experimental results prove the e‚ectiveness of our framework in the weakly-supervised seŠing. ACM Reference format: Meng ‹1, Xiang Ren2, Yu Zhang1, Jiawei Han1. 2016. Weakly-supervised Relation Extraction by PaŠern-enhanced Embedding Learning. In Proceedings of , , , 10 pages. DOI: 10.1145/nnnnnnn.nnnnnnn