Improve date extraction accuracy

Part of this task is to accurately define what exactly we want to extract date-wise. For example, if we extract the date from the URL, it might not actually be the date when the document was most recently adopted. But what if that's the only option? is that good enough? Or should we have no date at all? 

