Remote supervision relation extraction method and system based on relation hierarchy interaction

Abstract

The invention discloses a remote supervision relation extraction method and system based on relation hierarchy interaction, and the method comprises the steps: fusing three types of inputs: word embedding, relative position embedding and head and tail entity embedding through an information processing mechanism, and obtaining word embedding representation; coding the word embedding representation through neural network coding, and obtaining sentence representation; establishing a relationship hierarchical interaction structure, and obtaining a relationship-enhanced sentence representation; eliminating mistakenly labeled sentence instances, and obtaining sentence packet representation; and constructing a classifier through a multi-layer perceptron and a softmax activation function, obtaining a probability score of the sentence packet for a relation category, and performing relation extraction according to the probability score. According to the method, for the error tag problem and the long-tail distribution problem of the remote supervision relationship, the hierarchical structure of the relationship in a knowledge base is utilized to model the interaction relationship between the relationship hierarchies, more valuable clues are provided for a relationship extraction classification task, and the performance of a relationship extraction model is improved.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, and more specifically relates to a remote supervision relation extraction method and system based on relational hierarchical interaction. Background technique [0002] Relation extraction is an important subtask in the field of natural language processing. It aims to identify the semantic relationship between entity pairs in a given sentence. The extracted semantic relationship can be applied to downstream tasks such as knowledge graph construction and question answering systems. For example, in the given sentence “[Sergey] e2 is one of the co-founders of [Google] e1.”, the purpose of relation extraction is to identify the relationship between “Google” and “Sergey” as the “founder” and represented as a triple <Google, founder, Sergey>. In the knowledge base, the representation form of each piece of knowledge can be understood as this triplet. [0003] Tradition...

AI Technical Summary

Problems solved by technology

[0012] In view of this, the present invention provides a remote supervision relationship extraction method and system based on relationship hierarchy interaction, which uses the hierarchical structure of the relationship in the knowledge base to model the interaction relationship between each relationship level, and solves the wrong label of the remote supervision paradigm. problem and the long-tail distribution problem

Images