Diseño e implementación de un modelo de aprendizaje automático que ayude a extraer frases de apoyo para etiquetas de opinión provenientes de tweets.

Abstract

Abstract: Due to the amount of information circulating on the Twitter social network, it is difficult to know the sentiment behind a tweet. Artificial neural network models can analyze the information available from tweets and help capture sentiment of a specific tweet that may, for example, affect the brand of a company or devastate earnings because it gives a negative tone. In this work, it is proposed application of a neural network model to predict the polarity of a tweet in function of the phrases extracted from the text of the tweet. The TensorFlow framework is used to build and evaluate the neural network model using a data set tagged from Kaggle's competitor. The quality of the data set is validated through an analysis carried out by a Psychology professional. The results show that the Sequential Recurrent Neural Networks algorithm presents a better performance, achieving 87% accuracy. The performance and performance of the algorithm is visualized through a web application.