This looks really cool and is badly needed. Our company would kill for a PDF to semantic HTML algorithm (or service) too, using machine learning based on comput

This looks really cool and is badly needed. Our company would kill for a PDF to semantic HTML algorithm (or service) too, using machine learning based on computer vision. Existing options just vomit enough CSS to match the PDF output, rather than mark up into headings, tables and the like.