IIRC their first experiments focused on the names of keywords and types, where they found, e.g. that "text" is a more familiar type name than "string". Presumably they'll later apply the same approach to syntax and then semantics.