Abstract generation method and system applied to reading robot

Abstract

The invention discloses an abstract generation method and system applied to a reading robot. The method comprises the following steps that 1, a candidate abstract extraction unit extracts an abstractof an article most similar to a current article in a training set to serve as a candidate abstract; 2, a framework selection unit for screening out important frameworks in the article by using two measurement methods of a framework relationship and framework similarity; 3, a frame coding unit acquires vector representations of frameworks; 4, an article encoding unit acquires vector representationsof articles; 5, a framework and article interaction unit calculates the mutual influence between the frameworks and the articles by the frameworks and article interaction units to obtain final article representations; 6, a decoding unit is used for generating an abstract according to the obtained article representations. The framework semantic information is introduced for the first time and fused into the abstract generation model, the mutual influence between the frameworks and the articles is calculated, and the abstract generation result is effectively improved. The method is clear in thought structure, obvious in effect and high in expandability.

Description

technical field [0001] The invention belongs to the field of natural language processing research, and in particular relates to a method and system for generating abstracts applied to reading robots. Background technique [0002] An abstract is a simple coherent essay that comprehensively and accurately reflects the main content of a text. Automatic summarization is the use of computers to automatically extract summaries from original texts, aiming to provide users with short representations of texts and facilitate users to obtain text information quickly and efficiently. This technology is widely used in various fields, such as military, public safety, enterprise, publishing, medical and so on. At the same time, facing the explosive growth of data, text summarization has gradually become a hot spot in natural language processing (NLP). [0003] In the early days of text summarization technology, there were mainly some feature-based methods. Knight, K.; and Marcu, D. 2002...

AI Technical Summary

Problems solved by technology

[0006] Although a lot of research has been done and great progress has been made, the above work either directly splices external knowledge as word vector features into the text word vector, or extracts the entity information in the article to guide the summary generation, and The complex semantic information contained in the article itself is not used

Images