Incorporating pre-training in long short-term memory networks for tweet classification

The paper presents deep learning models for tweet classification. Our approach is based on the long short-term memory (LSTM) recurrent neural network and hence expects to be able to capture long-term dependencies among words. We first focus on binary classification task. The basic model, called LSTM-TC, takes word embeddings as inputs, uses LSTM to derive the semantic tweet representation, and applies logistic regression to predict the tweet label. The basic LSTM-TC model, like other deep learning models, requires a large amount of well-labeled training data to achieve good performance. To address this challenge, we further develop an improved model, called LSTM-TC*, that incorporates a large amount of weakly labeled data for classifying tweets. Finally, we extend the models, called LSTM-Multi-TC and LSTM-Multi-TC*, to multiclass classification task. We present two approaches of constructing the weakly labeled data. One is based on hashtag information and the other is based on the prediction output of a traditional classifier that does not need a large amount of well-labeled training data. Our LSTM-TC* and LSTM-Multi-TC* models first learn tweet representation based on the weakly labeled data, and then train the classifiers based on the small amount of well-labeled data. Experimental results show that: (1) the proposed methods can be successfully used for tweet classification and outperform existing state-of-the-art methods; (2) pre-training tweet representations, which utilizes weakly labeled tweets, can significantly improve the accuracy of tweet classification.