Desarrollo de un modelo de síntesis y ranking de hechos relevantes aplicado a medios de prensa digitales en base a minería de datos y retroalimentación del usuario final.

Abstract

Internet users’ growth, which has been increased even more during the COVID-19 pandemic, along with the exponential growth of electronic news, has generated the need to create new tools which allow analyzing and synthesizing the ideas which want to be transmitted. In this project a model for news clustering, synthesis of relevant facts and ranking is presented. The model is based on data mining practices using large language models (LLM) and machine learning (ML), and allows feedback based on the end user's assessment. The model, based on several factors, creates a composite metric that allows filtering the information that is considered useful and valuable knowledge for the end user. The development of this model has been carried out following the reference framework of the Cross Industry Standard Process for Data Mining methodology. The CRISP-DM methodology has allowed gradual progress while being able to refine the results based on the input and output variables of each of the phases followed. Obtained results after implementing the developed models have been satisfactory. A total of 7,761 clusters of related news stories have been generated, and the relevant content for the end user has been synthesized. It can be deduced from the analysis that the average of the content after being scored is 54.13 in a scale of 100 points. At the same time, a regression model was built to predict the authority that the news story would generate on the Internet, the prediction score obtained has been of 91.85%.