This is cool, how can I go about using this for my own dataset - .pdf, .html files etc?