A News Keyword Extraction Method Considering both Semantics and Pragmatics

Based on pragmatics theory, this study aims to incorporate news users' information into keyword extraction. It integrates the popularity of news keywords among users into the TextRank model to improve the effectiveness of single-document keyword extraction. Linguistic theories are employed to explore the role of pragmatics in news keyword extraction. The popularity of candidate keywords among users is integrated into TextRank, constructing a candidate keyword graph and calculating the probability transition matrix. Through iterative calculations of candidate keyword scores, the top N words are selected to generate news keywords. It is necessary to incorporate news users' information into news keyword extraction based on pragmatics theory. The introduction of keyword popularity effectively retains the semantic and pragmatic information of news texts. Testing with datasets that evaluate semantic information, the Hot-TextRank algorithm outperforms the comparative methods in different news domains when using Top N=3 and Top N=5. With Top N=3, the average F-score improvement across various domains is 2.2%. Additionally, experimental results demonstrate that the Hot-TextRank algorithm can retain the pragmatics information of news texts and has the ability to extract appropriate keywords based on user characteristics. This paper innovatively proposes theHot-TextRank algorithm. The algorithm introduces the popularity of news keywords in users into keyword extraction, which provides technical support for news keyword extraction considering user information.