Tweet Act Classification : A Deep Learning based Classifier for Recognizing Speech Acts in Twitter

Speech act provides valuable insight for understanding the communicative intention and behaviour of a user utterance. This applies for conversation on any platform which includes social media network such as Twitter. This paper presents a deep learning based tweet act classifier (speech acts for Twitter) for understanding the content and intention of tweets and for discovering the valuable communications amongst the tweeters. A generic taxonomy of seven speech acts has been created for the task. A Convolution Neural Network (CNN) based Deep Learning (DL) architecture is integrated with different classifiers at the final layer such as softmax or linear Support Vector Machine (SVM) to develop a tweet act classifier. To make the classifier even more robust, various hand-crafted features are added to the system to boost its effectiveness towards improving the performance of the system. Experimental results indicate that the proposed model achieved good results with an average F1-score of 0.71 and outperformed state of the art approaches.