If those files are in different directories, you could use the Process Documents from Files operator. This way you can tag each directory with a label so that when you build a model (i.e. Naive Bayes) you could see how well specific documents classify.
Since they are in a XLS links, you could use Get Pages operator in conjunction with a loop to extract each URL, get the page, and save it.
You could append them with the macro %{t} which will give you a timestamp. Then you'd have to build a process that converts that timestamp to say 2017-05-16, which you can then aggregate the documents on.
