A joint entity-relationship extraction method and system based on attention mechanism

Abstract

The invention relates to an attention mechanism-based entity-relationship joint extraction method and system. The steps of the method include: converting the triplets of entities and relationships marked in the training data into a form in which each word corresponds to a predefined type of label; mapping each word in the sentence of the training data into a corresponding word Vector, input the neural network model based on the attention mechanism, and train through the backpropagation algorithm to obtain the label prediction model; input the sentence that needs to be extracted from the entity relationship into the trained label prediction model, and predict the label corresponding to each word , according to the corresponding relationship between the label and each word in the triplet, the entity-relationship triplet existing in the sentence is obtained. The system includes a preprocessing module, a model training module and a result processing module. The present invention improves the performance of joint extraction of relational entities through more effective use of key information in sentences, and has good practicability.

Description

technical field [0001] The present invention relates to deep learning and natural language processing technology, in particular to an attention mechanism-based entity-relationship joint extraction method and system. Background technique [0002] In recent years, with the rapid development of Internet information technology, news, social networking and other websites generate massive amounts of new data every day. These data contain a variety of contents, including a lot of very valuable information, which plays a vital role in people's lives. In order to extract and effectively use these valuable information, the concept of knowledge graph is proposed. In the knowledge graph, special nouns such as names of people and places in massive data are represented as entities, and the connection between any two entities is represented as a relationship. Such massive data is represented as a triplet of entities and relationships (entity 1, relationship, entity 2). Although the exis...

AI Technical Summary

Problems solved by technology

Although this type of method uses a neural network to predict the label sequence, it does not distinguish the importance of the words in the sentence from the currently predicted words, so that it cannot make better use of the information of related words while ignoring the irrelevant words. information

Images