Criminal Text Corpora

Question

My goal is to bolster a database with charge type categories and decipher between, criminals, victims, interested parties, and suspects. I have searched for a corpora of text that may already have been classified regarding crimes, charges, victims, and other criminal activity. I have started building a set but this will be quite small compared to any available corpus. Basically, is there a corpus / are there corpora related to criminal activity and direct charge types such as murder, homicide, and drugs?

Answer 1

For historical data on crime, I suggest the Old Bailey Corpus (comprising the time from 1674 to 1913). It exhibits courtroom language and is as close to spoken language as we can get for this time period. It has also a great detail of annotation of the different speakers (role, class, etc.)

Answer 2

I don't know exactly what you mean by "related to criminal activity". There are huge corpora of judicial decisions that relate to criminal activity, and recently you can even get recordings of arguments and questions from SCOTUS. Such data "relates", rather remotely. Trial transcripts are not a good source from a research-useful perspective (I'm assuming you don't have infinitely deep pockets and infinite time to figure out what cases you care about). A trial transcript only contains a minute piece of existing evidence (such as the perpetrator saying "I took care of that", but not the entirety of what was recorded). Some amount of that data exists as recordings (assuming that conversations of criminal activity was recorded), but was mostly not transcribed (and almost none of that data is out in the wild).

Another version of "related to criminal activity" would have to do with police arresting and interrogating suspects. Again, there are not publicly available transcripts of such interactions generally, though snippets can appear in court transcripts and thus appeals. With the increase in police body-cam recordings you might expect more such data to make its way out into the wild, but again there is the countervaling privacy concern, so I would not count on there ever being any such source of data.

Roger Shuy's publications are probably the best source of such data. If you want examples of particular kinds of data, it might be possible to find an example, but all such data is extremely opportunistic and the antithesis of a random sample, which is often a desideratum for linguistic databases.