Word Representation Learning Based on Bidirectional GRUs With Drop Loss for Sentiment Classification

Sentiment classification is a fundamental task in many natural language processing applications. Neural networks have achieved great successes on the sentiment classification task in recent years, since recurrent neural networks and long-short-term memory networks have the ability to deal with sequences of different lengths and to capture contextual semantic information. However, the effectiveness...