Deep learning and distributional semantic model for Indonesian tweet categorization
Yanfa Adi Putra, Masayu Leylia Khodra · 2016
Twitter is one of the most well known social media in this era. Each day, many of this tweets are tweeted by people. In this experiments, Indonesian tweet will be extracted as set of features and will be categorized using machine learning. This experiment will conduct research to get the best configuration to categorize Indonesian tweet. To get the best configuration, this experiment will use lexical and distribution semantic model as feature extraction method. These two methods will later be compared to get the best one for Indonesian tweet representation. In addition to feature extraction methods, to get the model with the best accuracy, this experiment will also compare: SVM, ANN, and DNN. The best accuracy obtained by using ANN with word2vec CBOW at 82.94% for testing using 10-fold cross-validation and obtain 77.98% accuracy with data testing using deep neural network algorithm.