Construcción de un corpus de gran escala en el idioma español cuyos documentos reflejen opiniones respecto a productos textiles

Abstract

Currently, there is a boom in introducing Machine Learning models to various aspects of everyday life. A relevant field consists of Natural Language Processing (NLP) that seeks to model human language. The difficulty of training models to learn a language is high. A key and basic component for these intelligences to learn properly consists of the data, which in the case of NLP, is mostly in English. This research project arises from the problem of finding large-scale inputs, in languages other than English, to feed Deep Learning models that produce texts automatically. Four main results have been generated: 1) A methodology to build a large-scale corpus, easily scalable to different domains and languages, 2) A corpus in Spanish, within the domain of comments on textile products, with more than 170 thousand documents that obtained good results from human and automatic evaluations, 3) A computational system that automated the construction of the corpus from beginning to end, from the collection of documents to their evaluation, and 4) baseline results of a generational model that serve as a point of reference for future research within the automatic generation of texts within the textile domain