RNI does not require the entity type for matching, so you can enter anything you want in the relevant field (using the NameBuilder class). For example, if you want to assign the entity type 'World Leaders', you can do so. Then, if a query specifies the 'World Leaders' type, it will narrow results to entries of that type.
For matching, RNI works best on names of people. Within a single language, it should do nearly as well on other sorts of names as well. For instance, it should do fine with matching name variations on 'Scarborough Faire'.
It gets tricky when you're working cross-lingually on an entity whose name should be translated rather than transliterated. For example, let's say you're matching Indonesian text to an English index, and you encounter the named entity, 'Bandara Nasional Reagan' (Reagan National Airport). RNI will, of course, get 'Reagan' right, and should do fine with 'Nasional'. However, it will not get 'Airport' from 'Bandara'.
Of course, the index, itself, can be multilingual. If it contains both 'Reagan National Airport' (English) and 'Bandara Nasional Reagan' (Indonesian), linked as different labels for the same entity, then RNI will do just fine.