I'm not sure "sign post" questions are appropriate however that is what this is. I am looking for a tool, a means of answering questions such as "Word X is used more by demographic Y than demographic Z." to achieve this end I think the following is required: A large diverse collection of documents and sources with meta-data which can then be used as selection criteria to define a corpus.
An example of such selection might be: Include/Exclude social media from x,y,z sources. Include/Exclude academic papers. Include/Exclude magazines (by name), newspapers (by name), blogs. Only select where the authors geographic zone is known or self identified, and restricted to some geographic zone(City, country). Restrict search by gender. Use audio to text capture from media sources? Only include books found under the following sections: a,b,c. etc.
Then apply word frequency counts by methodology (first x words from text), sum boolean word occurrence by document, in the case of audio-to-text indicate audio sample size per count. etc.
Does such a tool exist?