A Study on Recursive Neural Network Based Sentiment Classification of Sina Weibo

摘要

Analyzing sentiment hidden in Sina Weibo's huge amount of information can benefit online marketing, branding, customer relationship management and monitoring public opinions. In this paper, we show how a recursive neural network can be trained to classify Sina Weibo messages' sentiment. Considering syntactic and semantic meaning of the sentence, this method is much superior to just basing on sentiment dictionary. Extensive experiments on huge dataset of Sina Weibo demonstrate that this model consistently outperforms existing sentiment classification model on identifying hidden or implied sentiment.