Relation Extraction: A Brief Survey on Deep Neural Network Based Methods

Abstract

Knowledge, data, algorithms and computing power are the foundations of artificial intelligence (AI), for which knowledge is the most powerful support. An effective way to acquire knowledge, called Relation Extraction (RE), is beneficial to knowledge graph completion, path inference, logical rule reasoning, and many other AI tasks. In this survey, we provide a comprehensive review on deep neural networks (DNNs) based RE covering some main research topics about: 1) general framework, 2) supervised and distant supervised RE, 3) future research directions. General framework provides a full-view summarization on traditional methods, main component and basic conceptions of DNN-based RE. For supervised and distant supervised RE, this part summarizes DNN-based RE methods into a new taxonomies on these topics. In the end, to facilitate future research on RE, we also discuss some obstacles.