Aicyber's system for IALP 2016 shared task: Character-enhanced word vectors and Boosted Neural Networks
Steven Du, Xi Zhang · 2016
This paper introduces Aicyber's system for IALP 2016 shared task, the Dimensional Sentiment Analysis of Chinese Words. The system is an ensemble of several boosted one layer neural networks, each one is trained on a different type of Chinese word vector. Our best system mainly use position-based character-enhanced word embedding and FastText as word vectors and achieve Mean Absolute Error 0.577(1st) in valence prediction and Pearson Correlation Coefficient 0.671(1st) in arousal prediction.