Authors: Ethan Anderson, Taran Kavuru, Preethika Yetukuri, Song Liao, Jeffrey Young, and Dr. Long Cheng
Faculty Mentor: Dr. Long Cheng
College: College of Engineering, Computing, and Applied Sciences
Voice assistants have grown rapidly in popularity. They use voice recognition to understand their user's commands and carry out, in the case of Amazon’s Alexa, skills, which are auxiliary programs developed by a third party to add more functionality. These skills can be enabled by Alexa users from the skill store. Users are also able to leave reviews about the skill. Being able to quickly analyze these reviews enables a better understanding of the skill ecosystem and the problems being faced collectively across skills by users. One of these problems is when Alexa is unable to understand the user’s commands, leading to frustrating interactions and a subsequent negative review, which we classify as an Alexa Understanding Review (AUR). Developing a methodology for detecting these reviews and applying this methodology to a large review data set has shown that about 0.6% of all reviews and 1.2% of all negative reviews can be categorized as an AUR.
