LIA @ DEFT’2017 : Multi-view Ensemble of Convolutional Neural Network
Mickael Rouvier, Pierre-Michel Bousquet
Abstract : LIA @ DEFT’2017 : Multi-view Ensemble of Convolutional Neural Network This paper describes the system developed at LIA for the DEFT-2017 evaluation campaign. The goal of this evaluation campaign was to identify opinion expressed by its author on tweets. An ensemble of Convolutional Neural Network (CNN) was developped, where we varied the network parameters and initialized the input of the CNNs with different sets of word embeddings. The final system is a score-fusion. The system is ranked 1th at DEFT-2017.
Keywords : convolutional neural network, word embedding, sentiment embedding.