Modelo de estimación del impacto de los titulares de noticias publicadas en Facebook en la audiencia ecuatoriana basado en minería de texto.

Abstract

The emergence of social networks and their public access have allowed the criteria issued by groups or organizations to be received and interpreted by different types of audiences, which may affect their perception of reality. Depending on the interest, ethics and professionalism of the author, the message can have an impact on the social environment by redefining facts, truths, or beliefs and even more if it is about news of social interest. In this work, a supervised machine learning model is proposed to analyze and identify the sentiment transmitted in the news headlines published on Facebook by the main Ecuadorian newspapers, as well as their corresponding articles, and then estimate their impact on the local audience. The results show that the main newspapers in Ecuador meet the principle of neutrality in the publication of headlines on Facebook. However, their articles express defined positive and negative sentiments, which results in the fact that the contextualization of the words used in the publication influences the sentiments that it transmits due to the different meanings that these words may have.