A Survey Deep Learning Based Relation Extraction

Abstract

From human understanding of natural language to machine understanding of natural language, NLP has become the mainstream technology in the world.Extracting useful information from massive information, namely information extraction (IE), such as relation extraction (RE), is one of the important semantic processing tasks.With the explosion of web texts and the emergence of new relations, the margin of human knowledge has increased dramatically, and unstructured text processing has become a problem that needs to be overcome, so people have been working on RE for many years. From the earliest pattern matching to the current popular neural network, RE has made significant progress. After reviewing the development history of relation extraction, this paper analyzes and discuss the development of existing neural network models and pre-trained language models from static technology to dynamic technology and reinforcement learning; finally, combined with the latest development of NLP research technology, the future research direction and trend of entity relation extraction are prospected.