This paper presents Senti17 system which uses ten convolutional neuralnetworks (ConvNet) to assign a sentiment label to a tweet. The network consistsof a convolutional layer followed by a fully-connected layer and a Softmax ontop. Ten instances of this network are initialized with the same wordembeddings as inputs but with different initializations for the networkweights. We combine the results of all instances by selecting the sentimentlabel given by the majority of the ten voters. This system is ranked fourth inSemEval-2017 Task4 over 38 systems with 67.4%
展开▼