Word2vec and Clustering based Twitter Sentiment Analysis

ÖzetçeHigh dimensionality of feature space is major problem due to the sentiment analysis is usually considered as text classification problem. In this study, we investigated the applicability of “word2vec and clustering based text representation” method for Twitter sentiment analysis. We conducted experiments on two different datasets that are comprised of Turkish Twitter feeds from which one is subject-dependent and the other one is subject-independent. In classification phase, we utilized Support Vector Machine (SVM) algorithm. Experimental results show that the W2VC has been quite successful and has provided a tremendous advantage in terms of time and performance as it reduces feature space, but it does not provide enough success in terms of accuracy.