Unsupervised Learning Based Brand Sentiment Mining using Lexicon Approaches A Study on Amazon Alexa

Abstract

Consumer sentiment analysis has gained immense attention in the recent past. The abundance of data in today’s world, especially those generated from the social media platforms, has triggered sentiment exploration like never before. The analysis of consumer sentiments have indeed helped organizations in effective decision making worldwide. In the communication technology domain, voice activated virtual assistants (VAVAs) are one of the latest entrants and they are gaining immense popularity by the time. Brand sentiment studies on VAVAs being limited in number creates an opportunity to explore further. This study fits into the domain of sentiment mining and the purpose of the paper is to review the consumer sentiment towards the global leader brand in the voice activated virtual assistant product segment, Amazon Alexa. Of the various approaches available, the researchers chose unsupervised learning based lexicon approach to estimate the brand sentiment. Three popular lexicon based sentiment classifiers, TextBlob, VADER and AFINN, have been used in the present context for exploration purpose. To the best of the knowledge of the researchers, this research effort includes, for the first time, multiple lexicon based approaches in exploring the sentiment towards the brand Alexa. This study shows consumers to have a significantly positive sentiment towards the chosen brand. The output from the three comparative classifiers reveal similar results which also validates the robustness of the outcomes and that of the chosen methods. The study anticipates a bright sales potential of the brand. Also, the use of alternative lexicon approaches is expected to enrich the existing literature in the sentiment mining domain.