Corpus text-mining data only supports suggestive interpretations, not proof

Evidence

There are multiple limitations to this type of technology and methodology; simple errors in optical character recognition or automating metadata, for instance, can skew results wildly, not to mention larger issues of what types of sources are included or not included in any corpus. For my purposes, as a humanist rather than a scientist, I’m not looking to use the Rap Stats data to conclusively prove anything, but to isolate small patterns with which questions can be asked and broader interpretations can be suggested. — Andrew deWaard, Derivative Media, 336

Categorization

• QUESTION: How does the author’s caveat about ‘humanist rather than scientist’ position against the claim that ownership structure leaves fingerprints on artistic form? The reader aims to test causation vs. correlation and should track this methodological hedging closely.