The role of word and n-gram frequency analysis in inference of the content of scientific publication

Purpose: The paper presents an analysis of a scientific publication with regard to the frequency of words and n-grams. The research problem addressed was the question to what extent the text mining analysis of a scientific publication will allow to infer its content. Design/methodology/approach: The main research method is the analysis of tokenized text using word count functions, bigrams, and trigrams in selected sections of a scientific publication. The results of text mining analysis were compared with the classic, non-automated text analysis of the publication. The presented study is a pilot project in the form of a case study. Findings: The proposed method of analyzing a scientific text using an analysis of the frequency of words and n-grams enables inference of the content of the paper with regard to the names of variables involved in the study, the statistical apparatus used and the key literature cited. It should be observed, however, that the discussed method does not make it possible to establish which variables are moderators and which are mediators. Originality/value: In this paper, the text mining technique was used differently in the discussed study than in previous works. The publication was not examined in its entirety, as previous researchers did, but text mining analysis was applied to individual parts of the paper, i.e. the part discussing theoretical foundations of the research and the part presenting the research method, research results, and their discussion. This allowed for obtaining more precise results regarding the content of the publication.