Before we have developed a machine learning algorithm which can decide what contexts are involved in a corpus by checking which words are included, and in what order and within what subcontext?