What people wants
As everybody knows there are millions (namely 4.572.896) textual expressions emerged from French GDN (Grand débat national). It is not easy to make sense out of this amount of messages, written in a language which is often very different from the one used, for instance, by customers in the retail sector. Innoradiant has been charged to answer to two questions:
- Which kind of political figure or public officer people would like to get rid of?
- Which kind of privilege should be retired from which kind of political figure or public officer?
It is evident that questions like these cannot be answered by machine learning /AI techniques as they presuppose a full understanding of the sentences, not just tagging or categorizing them. It is a typical domain where ASI, thanks to its layers of cognitive human coded rules, can provide tremendously effective results (in terms of precision and recall).
It is worth noticing that this project was based on a special research track aiming at determining semantic classes of attended behaviors, i.e. “what people normally do”, via deep learning based clustering.
For more details on this project, please consult this post.