A multi domains short message sentiment classification using hybrid neural network architecture

Sentiment analysis of short texts is challenging because its limited context information. It becomes more to be done on resource language like Bahasa Indonesia. However, with various deep learning techniques, it can give pretty good accuracy. This paper explores several methods, such as multilayer perceptron (MLP), convolutional neural network (CNN), long short-term memory (LSTM), and builds combinations those three architectures. The architectures are intended get the best architecture models. MLP accommodates use previous model obtain classification output. CNN layer extracts word feature vector from text sequences. Subsequently, LSTM repetitively selects or discards sequences based their context. Those advantages useful for different domain datasets. experiments sentiment in Indonesia show that hybrid models better performance, same directly used another domain-specific dataset.