Polarity Classification for Target Phrases in Tweets: A Word2Vec Approach

Twitter is one of the most popular micro-blogging services on the web. The service allows sharing, interaction and collaboration via short, informal and often unstructured messages called tweets. Polarity classification of tweets refers to the task of assigning a positive or a negative sentiment to an entire tweet. Quite similar is predicting the polarity of a specific target phrase, for instance @Microsoft or #Linux, which is contained in the tweet. In this paper we present a Word2Vec approach to automatically predict the polarity of a target phrase in a tweet. In our classification setting, we thus do not have any polarity information but use only semantic information provided by a Word2Vec model trained on Twitter messages. To evaluate our feature representation approach, we apply well-established classification algorithms such as the Support Vector Machine and Naive Bayes. For the evaluation we used the Semeval 2016 Task #4 dataset. Our approach achieves F1-measures of up to similar to 90% for the positive class and similar to 54% for the negative class without using polarity information about single words.