A big issue in innovation research is to attribute patents to particular people. Patent disambiguation is an attempt at creating some sort of person ID, when the underlying data has no such information. One application is to find the same person inventing under the same name but at a different company (signalling that he has changed jobs) and so on. In doing analyses like these that disambiguation becomes essential. There is no "definition", everyone tries to come up with their own algorithm to unify S. Smith, Microsoft Corp from Steven Smith, Microsoft Corp, there there are always false positives.
This is quite cool. What is the difference between raw and disambiguated? I have an educated guess, but I would appreciate a definition.