A method and system for automatically generating summaries that integrate semantic scenes

Abstract

The invention belongs to the field of natural language processing research, and specifically relates to a method and system for automatically generating abstracts that integrate semantic scenes. The specific content is as follows: 1. The graph construction unit, which constructs the semantic scene graph and the word relationship graph respectively according to the framework in the article; 2. The article encoding unit, which uses the pre-trained model to obtain the vector representation of the article; 3. The graph encoding unit, which uses the graph model Obtain the vector representations of the semantic scene graph and the word relationship graph respectively; 4. The graph interaction unit calculates the influence of the word graph on the semantic scene graph and obtains the graph vector representation; 5. The feature fusion unit fuses the article and the graph vector representation to obtain the final The article vector representation; Sixth, the decoding unit, which inputs the article representation into the decoding unit to obtain the final summary. The method of the invention respectively constructs a semantic scene graph and a word relationship graph from the structure and word levels, and calculates the relationship between the two to guide the generation of abstracts, has strong scalability, and can effectively improve the quality of generated abstracts.

Description

A method and system for automatic abstract generation based on semantic scene fusion technical field The invention belongs to the research field of natural language processing, and is specifically a method for automatically generating abstracts of a fusion semantic scene. laws and systems. Background technique [0002] The automatic generation of the abstract refers to the use of a computer to generate a brief and important information from the original text on the basis of the original text. message text. With the advent of the era of big data, people need a lot of time to obtain important information from massive texts, and Automatic text summarization is becoming more and more important as a method to improve the efficiency of obtaining important information. [0003] With the continuous development of deep learning, methods based on deep learning are widely used in summary generation tasks, Its model architecture mainly uses sequence-to-sequence encoding, includin...

AI Technical Summary

Problems solved by technology

Although the sequence-to-sequence-based deep learning method has achieved great success, the model architecture tends to capture the contextual information of the text, ignoring the inherent structural information of the text, which in turn affects the performance of the model

In addition, current graph-based deep learning methods only consider the relationship between words in the text or the structural information of the text, while ignoring the relationship between the two

Images