Generación de un corpus para detección de competidores en el idioma español mediante minería de opiniones comparativas. Caso de estudio: sector textil en la provincia del Azuay

Abstract

Currently, with the advancement of technology and even more so with the arrival of the pandemic, the use of digital platforms has increased. A study presented by the Ecuadorian Chamber of Electronic Commerce for the year 2020 shows that electronic commerce has increased the use of online digital platforms by at least 15 times compared to 2019 with the arrival of the pandemic. Due to this, companies to do market research must look for new sources of information. Therefore, the internet has become an intangible input for any business strategy. A fundamental part of a commercial strategy is to analyze the competition, this analysis in previous years according to the literature was generally carried out through surveys, but with the arrival of digital platforms this method has changed and today the data can be extracted from the web to then implement a Competitive Intelligence (CI) process, which allows a complete analysis to have a competitive advantage. CI comprises several steps, this research addresses all these steps, but focuses mainly on the initial step, data collection and data analysis, which is a fundamental step for CI, where there are currently problems such as: lack of corpus in Spanish specialized for CI, so researchers do not have the facility to implement machine learning models that help them to have a competitive advantage. This research presents a methodology for the creation of a corpus in the Spanish language that allows algorithms to be trained in order to detect competitors in the context of the textile sector. Two main results have been generated: 1) A methodology using text mining techniques (comparative opinion mining and named entity recognition) to build a corpus focused on Competitive Intelligence. 2) A corpus in Spanish, within the domain of social network comments, which serves as a basis for future research related to competitive intelligence, specifically in the detection of competitors in the Spanish language, where the CI was strictly restricted by the lack of a corpus. Finally, the usefulness of the corpus developed has been evaluated through a Dashboard created based on a case study carried out in the context of the textile sector in social networks. It has been shown that it is indeed useful for the textile sector, however, it is recommended to carry out a new validation with companies that are directly related to the textile sector and thus obtain a more direct validation, it is also recommended to evaluate in other sectors.