Bert-based multi-layer attention mechanism relationship extraction method

Abstract

The invention relates to an entity relationship extraction method, and in particular relates to a bert-based multi-layer attention mechanism relationship extraction method. The method comprises the following steps of S1 acquiring sample data; S2 dividing the sample data; S3 replacing entities in the sample data with #; S4 obtaining an entity and connecting the entity with the sentence through $ to form a training sample; S5 inputting the training sample into a BERT language model and a full connection layer, and carrying out vectorization processing to obtain a fusion feature vectorization representation; S6 sequentially adding word attention mechanisms to the training samples, and adding weights to local features of the training samples; S7 adding a sentence-level attention mechanism, and adding a weight to the global feature; and S8 inputting the sentence-level feature vector into a classifier after passing through the full connection layer, and obtaining a classification result.

Description

【Technical field】 [0001] The invention relates to an entity relationship extraction method, in particular to a bert-based multi-layer attention mechanism relationship extraction method. 【Background technique】 [0002] In the natural language processing technology center, the construction of knowledge graph plays a very important role in the development of artificial intelligence. In the construction of knowledge graph, knowledge extraction is the core foundation. Knowledge extraction mainly includes three sub-tasks: entity extraction and relationship extraction. , Event extraction, and entity relationship is the core task and important link of knowledge extraction. The main goal of entity relationship extraction is to identify and determine the specific relationship between entity pairs from natural language text, which provides basic support for intelligent retrieval, semantic analysis, etc., and helps to improve search efficiency. [0003] The supervised relationship extr...

AI Technical Summary

Problems solved by technology

[0003] The supervised relationship extraction method is currently considered to be a method with better relationship extraction effect. It solves the relationship extraction task as a classification problem, designs effective features according to the training data, thereby learning the classification model, and then uses the trained model to predict relationship, most of the existing supervised relationship extraction models are based on the word2vec pre-training model, which cannot explain the polysemy of words in the text, and most current relationship extraction models cannot make full use of the local features of text information and global features

Images