A Comparative Study for Sentiment Analysis: LDA and LDA2Vec

Abstract

Internet technology developed a new era of data mining by sharing their reviews, thoughts and ideas regarding any topic. Web 2.0 emerged as a new generation of Internet Services and created various social medium like Twitter, Facebook and creative blogs a data warehouse for exchanging and expressing their views. Nowadays enormous amount of data is accessible linked to any topic whether it will be Political agenda, Business enterprise, Survey organizations or different advertising firms. Customer contentment is the major component for various survey companies so as to improve their services. LDA2Vec is new approach proposed by Chris Moody and much research related to this approach is still not done. In this paper we used Latent Dirichet allocation (LDA) and LDA2Vec model for Sentiment classification. We evaluated performance of both the models by using corpus of 1000 records. After execution of model, we can easily demonstrate by our experimental results that hybrid approach of LDA2Vec (LDA and Word2Vec) performed better in comparison to LDA approach.