Benchmarking Sentiment Analysis Approaches on the Cloud
Richard Sinnott, Shucheng Cui · 2016
Social media resources such as Twitter provide global services for citizens to express opinions on people, products, events or even themselves. Often this data captures the mood (sentiment) of the tweeter. Accurate and timely extraction of sentiment from such big data can be used for many population-wide business and research scenarios. Whilst a range of sentiment analysis approaches has been taken, little systematic comparison of these approaches has been undertaken. The motivation of this paper is to investigate various sentiment analysis approaches and evaluate their accuracy and performance for Twitter-based sentiment analysis on major Cloud facilities across Australia. We consider especially the impact of training data on performance and accuracy of sentiment analysis. To support this, we present a Cloud-based architecture and its realization through an elastic, distributed, data processing system used for harvesting, analyzing and storing large-scale Twitter data sets.