Alexa, My Love: Analyzing Reviews of Amazon Echo

The phenomenal success of Amazon Echo as a voice-activated wireless speaker has intrigued us to understand how consumers use the device with an embodied conversational agent in real-world situations. In addition, we are interested in learning what kind of relationship people develop with the device, which has a humanized name. Thus, we performed a large-scale analysis of the user reviews of Amazon Echo, with the dataset consisting of 55,502 reviews spanning over 2 years from May 2015 to May 2017. We first conducted a qualitative study using content analysis of 144 representative reviews to discover what features people like and dislike about Echo. Two case studies were then performed, one for the hands-free feature that customers raved about and the other for speech recognition and understanding which is one key feature of the Echo. To automate the analyzing on a large scale of reviews, we used natural language processing techniques for sentiment analysis and compared Echo reviews with reviews of two traditional electronic devices, Fire Tablet and the popular Bluetooth speaker Dknight MagicBox II. Interestingly we found a significant number of reviewers personify Echo as an assistant, a friend, or a family member. Some reviewers even compare Echo with their wives or girlfriends. Using emotion analysis, we found the reviews personifying Echo showed more positive emotions than those simply treated Echo as a device. In the future work, we plan to conduct field studies to understand interesting use cases of Echo and the novel applications enabled by such powerful ubicomp devices.