Sentiment Weighted Word Embedding for Big Text Data

Abstract

Sentiment analysis is the practice of eliciting a sentiment orientation of people s opinions (i.e., positive, negative, and neutral) toward a specific entity. Word embedding techniques like Word2vec are effective approaches to encode text data into real-valued semantic feature vectors. However, they fail to preserve sentiment information that results in performance deterioration for sentiment analysis. Additionally, big-sized textual data consisting of large vocabulary and its associated feature vectors demand huge memory and computing power. To overcome these challenges, this research proposed a MapReducebased sentiment-weighted Word2Vec (MSW2V) that learns the sentiment and semantic feature vectors using sentiment dictionary and big textual data in a distributed MapReduce environment, where the memory and computing power of multiple computing nodes are integrated to accomplish the huge resource demand. Experimental results demonstrate the outperforming performance of the MSW2V compared to the existing distributed and non-distributed approaches.