Sentiment Analysis Model on Weather Related Tweets with Deep Neural Network

Abstract

Weather related tweets are user's comments about daily weather. We can gain useful information about how weather influence people's mood by analyzing them. This is what we called opinion mining in natural language processing field. Traditional opinion mining algorithm use feature engineering to build sentence model, and classifier like naive bayes is used for further classification. However, these feature vectors can sometimes be insufficient to represent the text, and they are manually designed, highly relevant to the problem's background. In this work1, we propose a method modeling text based on deep learning approach, which can automatically extract text feature. As for word's vector representation, we incorporate linguistic knowledge into word's representation, and use three different word representations in our model. The performance of the sentiment analysis system shows that our method is an efficient way analyzing user's sentiment on weather events.