What people desires
An market studies agency was conducting for a big customer surveys about certain factors building up the image of several car brands. They wanted to know which would have been the results if, rather than asking customers to answer to online surveys, they could capture the same factors as they were emerging from spontaneous, unsolicited conversations. ASI was then used to “convert” the questionnaire in a set of linguistic cartridges capturing those same factors in forum conversations. These are the results as a heatmap
Different factors (affordability, after sale service, comfort) are crossed with different brands. The more a cell is red, the worse the evaluation by user forums, the greener a cell is the better is the evaluation. For instance, Volkswagen is performing very badly on reliability (diesel gate) whereas Alfa is best performing on driving pleasure.
In this task it was central to work on emotions and to parametrize them according to the automotive domain. ASI was able to discover different emotions car owners felt about cars or aspects of a car. Thanks to the exploitation of self tuning strategies developed in years of research, it was also able to characterize them lexically in a non-trivial way: for instance by understanding that a word such as “aggressive”, normally negative, was denoting a positive emotion, close to enthusiasm when referred to a car look or driving style.