Lexical-semantic similarity in scientific research articles in Spanish: An approach to latent semantic analysis

This research focuses on lexicaL-semantic similarities found in three text variables (key words, abstract, and content in scientific research articles). Two scientific areas (biological sciences and social sciences) from the values of lexica I-semantic similarity are compared. The study employs a representative stratified sample of 22 scientific research articles in these two areas, which have been included in a corpus of 675 scientific articles. To determine lexicaL-semantic similarities among the variables, a computer-statistical method is employed, called Latent Semantic Analysis. The findings help assert, on the one hand, that the abstract 'macro-semanticizes' better the global semantic content of the article than do the key words. On the other hand, no meaningful differences among the averages of lexica I-semantic similarity in the scientific areas studied are revealed. These findings are accounted for in terms of the generally complex standardization processes which tend to homogenize the production of this type of articles.