Text understanding method based on external knowledge embedding

Abstract

The invention discloses a text understanding method based on external knowledge embedding. The method comprises the following steps: dividing entities in a data set through a clustering algorithm, andadding a negative triad in the data set through random sampling splicing; then, building a BERT network under a Keras deep learning framework, and coding entity relations in the knowledge graph to obtain a corresponding representation vector; extracting low-level features of different dimensions by using multi-granularity convolution so as to further improve the feature extraction capability of the model; mining the deep relationship between entities through the capsule neural network so as to complement the knowledge graph and improve the knowledge accuracy; fusing the code of the entity relationship serves as external knowledge with text entity information; and finally, interacting the knowledge vector with the text context by using a self-attention mechanism, and acquiring an answer through a multi-layer perceptron. Therefore, the method can enable a machine to well understand the meanings of the preceding and following texts, and enables the intelligent level of the network to behigher, and closer to the cognitive thinking of human beings.

Description

technical field [0001] The invention belongs to the technical fields of natural language processing, knowledge representation and machine reading comprehension, and in particular relates to a text comprehension method based on external knowledge embedding. Background technique [0002] The knowledge graph was officially proposed by Google in 2012. Its original intention is to optimize the search results and enhance the user's search quality and experience. The knowledge map provides a more effective way to represent, organize, manage and utilize massive, heterogeneous and dynamic big data on the Internet, making the network more intelligent and closer to human cognitive thinking. It is usually denoted as triplet G(E h ,R,E t ), where E is the entity set in the knowledge graph, E h and E t Respectively represent the head entity and the tail entity, and R represents the relationship between entities. [0003] It is very effective to use the pre-trained language model to s...

AI Technical Summary

Problems solved by technology

[0007] Although knowledge graphs have been widely used at present, not only in terms of data volume but also in the accuracy of data triplets, there are still several problems: one is the number of data triplets in different The knowledge map is uneven. With the continuous increase of knowledge every day, the knowledge map needs to be completed quickly; the second is that there are still inaccurate knowledge relationships in the existing knowledge map. Therefore, how to ensure the knowledge of the knowledge map is accurate is also a problem

Images