Abstract: In this paper, co-word analysis is applied to the text file to explore the semantic relevance of the text topic. Since text file is different from the science literature, andlacks keywords, the author uses the topic model to reduce the dimensionality of the text, and uses the topic words generated by LDA as the research object of co-word analysis. The experiment results show that the medium frequency words can better reflect the theme of the text. Therefore, the author uses Zipf’s law and the same frequency theory to select medium frequency words as the topic words set, calculates the semantic association strength with co-word analysis, and clusters the topic words using K-means clustering algorithm. The author then makes an experiment with the news text relevant to “innovation and entrepreneurship”, and concludes that this method can reflect the semantic relevance of the theme of the text better.

Key words: Topic model, Zipf’s law, Co-word analysis, Subject words clustering