Sentiment Analysis of YouTube Video Comments Using Deep Neural Networks

摘要

Over the years, social networks have become an important vehicle for communication. Many users on YouTube use comments to express opinions or critique a subject. The amount of comments, for famous videos and channels, is huge, which poses the challenge of analysing user opinions efficiently. This article proposes a sentiment analysis model of YouTube video comments, using a deep neural network. We employed an embedding layer to represent input text as a tensor, then we used a pair of convolutional layers to extract features and a fully connected layer to make the classification. The output of the neural network is the sentiment classification among negative, positive or neutral. Two videos were chosen and their comments were classified by our model, by an alternative statistical model and by humans. The human classification was considered to be 100% accurate. The results showed that our model achieves better accuracy than the statistical model, and the classification accuracy is in the range 60%-84%.