Relationship extraction method and device based on tree-type capsule network

Abstract

The embodiment of the invention provides a relationship extraction method and device based on a tree-type capsule network, relates to the field of character processing, and can improve the accuracy ofrelationship extraction. According to the technical scheme, the method comprises the steps that all words in a text to be recognized are encoded through a sequence encoder, a word vector matrix is obtained, and the word vector matrix comprises word vectors obtained after all the words are encoded; and splicing the embedded representations of the two specified entities in the to-be-identified textand the embedded representation of the to-be-identified text to serve as sequence sentence embedded representations. Wherein the phrase structure tree of the to-be-recognized text is encoded by utilizing the tree-type capsule network to obtain a structure sentence embedding representation, the phrase structure tree is used for representing a syntactic structure of the to-be-recognized text, and each leaf node of the phrase structure tree corresponds to one word of the to-be-recognized text. And determining a relationship category between the two specified entities according to the sequence sentence embedding representation and the structure sentence embedding representation.

Description

technical field [0001] The present invention relates to the technical field of word processing, in particular to a relation extraction method and device based on a tree-type capsule network. Background technique [0002] Relational extraction refers to extracting the relationship between two entities in a sentence, where entities refer to words that represent real things. Generally speaking, words in knowledge bases such as Wikipedia or Baidu Encyclopedia can be called for entities. For example: the entities in the sentence The winery includes a large garden are winery and garden. [0003] The current common relationship extraction method is: input the vectors of two entities in the sentence into the neural network model, and the neural network model predicts the relationship between the two entities. [0004] The relationship extraction method based on the neural network model depends on the representation quality of the instance, and due to the complexity of natural lang...

AI Technical Summary

Problems solved by technology

[0004] The relationship extraction method based on the neural network model depends on the representation quality of the instance, and due to the complexity of natural language, the semantics of the entity that the vector can represent is limited, so the representation quality of the instance represented by the vector is poor, making the use of the neural network model The accuracy of relation extraction is low

Images