In addition, we also think that the accuracy of extracting implicit
information from the public users information could also be improved
having a larger and more specific list of categories. This
could be done working directly with customized natural language
processing techniques instead of relying on a third party web
service.