Clasificador de Noticias usando Autoencoders
Gonzalo Farías, Sebastián Vergara, Ernesto Fábregas, Gabriel Hermosilla, S. Dormido-Canto, S. Dormido · 2018 IEEE International Conference on Automation/XXIII Congress of the Chilean Association of Automatic Control (ICA-ACCA) · 2018
This article presents a classification system for news with Deep Learning. With this tool the news are classified in the following categories: Sports, Politics, Economics, Show and Police. Also they receives an scope: Local (Valparaíso), National (Chile) and International (Rest of the World). The classifiers were built using a database with 542 news labeled with the previous criteria. The features were extracted with Autoencoders (AE) to train an Artificial Neural Network (ANN) of multiple classes Softmax (Softmax ANNs). Both stages were stacked following the concept of Deep Learning. The results with the data test (156 news) reach a success rate of 92.3% for the category classifier and 87.2% for the scope classifier. The general success rate for both, category and scope was 83.75%.