Paper keyword extraction system and method

Abstract

The invention relates to the field of artificial intelligence, in particular to a paper keyword extraction system which comprises a training set, a Jieba word segmentation device, a cleaning module, a weight calculation model, a frequency calculation model and an output model. The paper keyword extraction method comprises the following steps: S1, acquiring words of a paper body in a training set by using a word screening module to form a word training set; s2, performing word segmentation through a Jieba word segmentation device to obtain words of all word training sets appearing in the abstract, the text or the summary in the paper, and outputting key words through a cleaning module; s3, inputting a result of the step S2 into a weight calculation model for training; s4, inputting a result of the step S2 into a frequency calculation model for training; s5, inputting the output results of the steps S3 and S4 into an output model for training; s6, inputting the target paper into the weight calculation model, the frequency calculation model and the output model to obtain keywords; a mode of combining an unsupervised method and a supervised method is provided to acquire keywords of the papers.

Description

technical field [0001] The invention relates to the field of artificial intelligence, in particular to a paper keyword extraction system and a method thereof. Background technique [0002] A paper is divided into abstract, keywords, table of contents, text, acknowledgments, references and summary. Readers need to quickly find out the keywords of the paper when searching or work needs, although the papers are set with a keyword item, but this keyword is not accurate, and readers need to judge by themselves. Since keywords mainly appear in the abstract, text and summary, the existing technology usually uses the following two methods for extraction: [0003] (1) The unsupervised method does not require a manually generated and maintained vocabulary, nor does it require artificial standard corpus for training, which greatly increases the operating efficiency of the system and reduces labor costs. The TF algorithm is a An unsupervised statistical-based keyword extraction algorit...

AI Technical Summary

Problems solved by technology

[0003] (1) The unsupervised method does not require a manually generated and maintained vocabulary, nor does it require artificial standard corpus for training, which greatly increases the operating efficiency of the system and reduces labor costs. The TF algorithm is a An unsupervised statistical-based keyword extraction algorithm, used to evaluate the importance of a word in a document set to a document, and can also count the frequency of a word appearing in a document, if a word appears in a document If there are too many times, it means that the word expresses the article more strongly, but the disadvantage of using the unsupervised method is that the accuracy rate is not high;

[0004] (2) The supervised method trains the weight ratio, which can obtain higher precision and more accurate weights, so that the result can reduce the possibility of error, but the disadvantage is that the running speed is slow

Images