[3Xin2] Poster session 1

[3Xin2-33] A Citation Selection Method Based on Literature Feature Analysis to Support Writing Review Papers

〇Sho Kumamoto¹ (1.Tokushima University)

Keywords:text mining, machine learning, NLP, LLM

In this paper, we propose a semi-automatic, efficient, and accurate method for selecting citations for systematic reviews. The number of literature data that can be collected free of charge is limited in the field that is the target of the review in this study. In preliminary experiments, we conducted a topic analysis of abstracts of literature using topic modeling, but found that it was difficult to extract important descriptions related to citation criteria using this method. Based on this, we constructed a model to classify the articles selected as positive examples and those excluded as negative examples in the existing systematic review, and analyzed the unique expressions in the abstracts and text data of the positive examples using feature selection methods and interpretation methods of classification algorithms to examine the algorithm for extracting important descriptions. We then examine algorithms for extracting important descriptions by analyzing the unique expressions in the abstracts and text data of positive examples using feature selection methods and interpretation methods of classification algorithms.