Tweet Classification with Convolutional Neural Network

Tweet classification task is used to determine opinion of tweets. Such opinion is useful for making new strategy and taking right decision as per situation. Due to high speed and high availability of internet, large numbers of people are involving in social media to share their opinion towards any happening event like sport. It needs to analyze the behavior of people whether they are happy or unhappy towards the event. We consider convolutional neural network model, one of the deep learning approach for tweet classification. We used word embedding technique like word vector for text representation. We used APNews corpus as word embedding technique to give word vector called pre-trained word vector. On top of pre-trained word vector, we apply convolutional neural network to know the polarity of tweet. Here, we map each word of tweet to already pre-trained word vector of APNews corpus. We fetched the tweets from social website and performed preprocess for training and testing purpose. During training phase, we found 87.25% accuracy. During testing phase, we found 79% accuracy of proposed CNN model.