We know that 70% of our communication is non-verbal, while verbal communication adds another 7%. Yet, interpreting non-verbal communication by humans is a time-consuming and highly subjective process. For this presentation, we show how machine learning is making qualitative concept testing more efficient, more scalable, and more objective. We demonstrate how the latest cloud computing and machine learning technologies of emotion analytics and text mining were applied in the process of qualitative in-depth interviewing. The combination of the two methods help observe nuanced consumer responses that never before were capable of being observed and compared by humans.
This ambitious project involves the development of a powerful tool for the automatic analysis of 100% of the comments of our quantitative surveys always with client centricity in mind. Through text mining models, a machine learning technique applied to non-structured data, we have managed to classify instantaneously the type of comment, the subject and the sentiment. In this way, it is possible to prioritize the most critical items, identify points for improvement and analyze the feasibility of the suggestions offered by our customers.
Exhaustive manual coding of open-ended responses is still an everyday reality. Thanks to technological advances we nowadays have text analytics. The main models behind it can only unleash their power with longer text, not with short statements we often face in market research. I explored models used for analyzing Twitter and Co. and transferred them to open-end analysis, with promising results.