Method for abstracting abstract text based on graph knowledge and theme perception

Abstract

The invention discloses an abstract text abstracting method based on graph knowledge and theme perception. The invention provides a model based on BERT, a neural topic model and a graph neural network, and the model is called GTASum. At the input end of a document encoder, using BERT to obtain hidden word vectors of the document; at the input end of a topic encoder, obtaining a word-topic distribution vector of the document by using a neural topic model; inputting the two vectors into a graph neural network for training to obtain context contents fused with theme knowledge, and generating a text by using a decoder based on Transform; meanwhile, the condition normalized LN layer provided by the invention can cooperatively train a neural topic model and a decoder, and feature selection is effectively carried out. The result shows that the method has better robustness and adaptive ability.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and designs a method for generating text summaries, specifically an abstract text summarization method based on graph knowledge and topic perception, and a text summarization method based on pre-trained language models, neural topic models, and graph neural networks method. Background technique [0002] With the development of computer performance and large-scale language models, the task of natural language processing (NLP) has achieved significant development. As one of the core problems of natural language processing tasks, the summarization task aims to allow people to quickly grasp the important information in the text. Text summarization has been widely used in many fields, such as news, finance, conference and medical treatment. Currently, there are mainly two methods for summarization tasks: extractive methods and abstract methods. Extractive methods mainly copy imp...

AI Technical Summary

Problems solved by technology

However, this approach relies on external tools, which may result in semantically fragmented output

Wang and Liu et al. built word-sentence document graphs, but it is difficult to capture semantic-level relationships in this way. Therefore, how to effectively build documents into summarizable graphs is also a difficult problem.

Images