Is it possible to use a CSV file as the set of input documents (i.e., where each row in the CSV file represents a different document)? We have a dataset containing thousands of documents and it's not practical to have each of these as a separate text file.