Media Buzzword Analysis Integrated with Phrase Vectors and Topic Model

Abstract

Buzzword analysis is one of the important research contents of natural language processing, and the research results can provide technical support for public opinion analysis. The purpose of extracting media buzzwords is to analyze the rules and changes of language change within a range. The traditional word feature-based buzzword extraction had some problems, such as low accuracy and low coverage, and this paper proposes a media buzzword analysis based on the combination of phrase vector and topic model, the core idea is to integrate the semantic similarity features, and use visualization technology to more intuitively show the overall language change rules. Visualization analyses uses a large number of corpus statistics, calculate the distance between words, and then convert into similarity, through word similarity calculation to show the distribution relationship between different words, and finally quantitative perspective to analyze. Our model is better than the traditional system, and the research results can provide corpus and model support for subsequent research directions.