Document worth reading: “Word Embeddings for Sentiment Analysis: A Comprehensive Empirical Survey”

This work investigates the perform of issues like teaching method, teaching corpus measurement and thematic relevance of texts inside the effectivity of phrase embedding choices on sentiment analysis of tweets, observe lyrics, movie critiques and merchandise critiques. We moreover uncover specific teaching or post-processing methods that may be utilized to spice up the effectivity of phrase embeddings in certain duties or domains. Our empirical observations level out that fashions expert with multithematic texts that are huge and rich in vocabulary are the most effective in answering syntactic and semantic phrase analogy questions. We extra observe that have an effect on of thematic relevance is stronger on movie and cellphone critiques, nevertheless weaker on tweets and lyrics. These two later domains are further delicate to corpus measurement and training method, with Glove outperforming Word2vec. ‘Injecting’ additional intelligence from lexicons or producing sentiment specific phrase embeddings are two excellent choices for rising effectivity of phrase embedding choices. Word Embeddings for Sentiment Analysis: A Comprehensive Empirical Survey