Aplicación de técnicas de Machine Learning para hacer análisis de polaridad de sentimientos en texto para detectar tendencias de opinión en plataformas online

Abstract

The Internet has allowed millions of people to connect and generate interactions, as in social networks, which has generated a lot of unstructured information, which is difficult for a group of human beings to analyze, due to its large amount. In this work, Machine Learning techniques are applied to analyze the polarity of sentiment in Spanish language, of the comments of Twitter users about various topics. Sentiment polarity analysis allows you to analyze opinion trends quickly and automatically, allowing companies and organizations to have valuable information for decision-making. Recurrent Neural Networks are implemented, which are one of the methods that show the best results for sequence analysis, through the application of Deep Learning, which belongs to the field of Machine Learning and which, in addition, avoids the need to perform extraction of characteristics, which would require careful selection by language experts. Keras is used to program the model with tensorflow, and accuracy results are obtained very close to the most advanced systems in the state of the art. The model is trained with a Dataset of 49,444 sentences labeled with positive or negative, based on the TASS corpus