This is a fair point, but internal representation of the input can create drastic differences in how it's processed. Try writing an interpreter without tokenizing and parsing the text into some kind of AST. While this could be described as process, I think it proves that the internal representation of data matters.