The entire set of language data to be analyzed. More specifically, a corpus is a balanced collection of documents that should be representative of the documents an NLP solution will face in production, both in terms of content as well as distribution of topics and concepts.
