Document modeling classification method based on WSD hierarchical memory network

Abstract

The invention discloses a document modeling classification method based on a WSD hierarchical memory network. The method comprises the following steps of firstly, obtaining a sentence embedding matrixof the similar sentence texts based on the word vectors through a Bert algorithm to obtain the semantic information between words; secondly, mapping the sentences into the sentence embedding matrix space to obtain the vectorized representation of the sentences; and finally, inputting the sequence data of the sentence-divided document into a BiLSTM model, obtaining the attention weight of each sentence at the same time, obtaining the vectorized representation of the document, and reserving the internal semantic connection of the document. According to the method, the document modeling with thehighest accuracy can be effectively obtained, the hierarchical relationship of the word and sentence cascading is fully considered, the semantic relation in the document modeling is increased, and the document classification with the higher inter-class data similarity is more accurate.

Description

technical field [0001] The invention belongs to the technical field of natural language processing and document classification, and in particular relates to a document modeling and classification method based on a WSD hierarchical memory network. Background technique [0002] The document modeling and classification algorithm in the present invention has important functions and significance for traditional supervised document classification. In the past, when faced with the classification problem of text labels, researchers would choose to integrate the vector space model into text classification. This type of method requires the text to be mapped into the vector space for supervised training, so as to use the trained Classifiers classify unclassified texts, but in the process, a large number of features need to be manually designed, and the semantic connections within the text are ignored. Therefore, in order to discover the semantic relationship between historical documen...

AI Technical Summary

Problems solved by technology

However, traditional document modeling still has the following problems: 1. The vectorization of documents is realized by word embedding based on word frequency variance, but this method cannot make full use of the relationship between text semantics; 2. Using the attention network to model The training time is prolonged, and the structural information inside the document cannot be effectively used, and the accuracy of multi-label classification cannot meet the actual application requirements.

Images