Minería de texto y modelamiento de tópicos en artículos científicos digitales

Abstract

Abstract:The degree work presentsthe application oftopic modeling,to identify barriers for women in STEM (Science, Technology, Engineer, Mathematics); topic modeling is a technique that extracts a specified number of topics from an input data set. The corpus used was generated from a search in the Scopus database, from 2000 to 2020; in addition, the use of keywords(STEM, women STEM, STEM barriers, gender barriers, gender disparities)and logical connectors; a total of 141 abstracts of digital articles were worked on. Topics were extracted using the LDA (Latent DirichletAllocation) algorithm, which assumes that a document is a random mix of latent topics; furthermore, with the use of thematic coherence in severalexperiments, it was possible to define the optimal number of topics (coherence graph).As a result, gender bias, lack of motivation, stereotypes and the lack of insertion programs for STEM women were obtained; as the most representative topics, which refer to some of the barriers of women in STEM.A visualization of the resulting topics is provided, through several graphs such as: PyLDAvis, word cloud, frequency diagram, heat map, histogram, TreeMap, bar diagram about the importance of each topic, among others.