Identification Of The Author'S Idea Based On The Modified Textrank Method

Taking into account the significant rate of new information volume formation, including scientific, semantic analysis of the text continues to be a relevant area of research. The results of this paper on extraction the author's idea can be applied to identify features of intellectual plagiarism to promote academic integrity. To identify the author's idea as the main content of the text, the authors use semantic and graph-based methods. The paper proposes a method for identifying the author's idea based on the modified TextRank algorithm. This method takes into account the pronominal anaphoric connections between sentences, allows to form a more complete description of the semantic relationships between sentences in the text. An experiment was carried out on scientific texts in the Ukrainian language, which confirmed an increase in the number of semantic links between sentences in comparison with the simple TextRank algorithm, which affects the weight of sentences and their order in the abstract.