An Automatic Text Summary Extraction Method Based on Improved TextRank and TF-IDF

2019 International Conference on Advanced Electronic Materials, Computers and Materials Engineering

无线电电子学;计算机科学;材料科学

Guan, Xinxin^1 ; Li, Yeli^1 ; Zeng, Qingtao^1 ; Zhou, Chufeng^1

School of Information Engineering, Beijing Institute of Graphic Communication, Beijing

102600, China^1

关键词: A-weighting;    Extraction method;    Frequency factors;    Hotspots;    Information explosion;    NAtural language processing;    Part Of Speech;    Text abstraction;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/563/4/042015/pdf
DOI  :  10.1088/1757-899X/563/4/042015

来源: IOP

【 摘 要 】

In the era of information explosion, more and more people advocate "fast reading". Compared with some texts longer texts can express vivid images but it will waste time. Therefore, automatic text abstraction has become one of the research hotspots in natural language processing. In order to ensure the accuracy of automatic text abstraction by computer, a weighting method based on comment factor, position factor and weighting of part-of-speech factor is designed which mainly combines and improves TextRank and TF-IDF algorithm. This method focuses on introducing the reader's comments as a comment factor, supplementing the shortcomings of paying too much attention to the original text and ignoring the reader's comments. Secondly, the frequency factor and part-of-speech factor are introduced to ensure the readability of the generated abstracts. The experimental results show that the method has good improvement in accuracy, consistency and connectivity.