The content processing could be good, especially if it was to decrease the complexity of the language and simplify the concepts. I would take care that the descriptions could be horribly wrong and misleading.

I would probably look to monetize in a different way:

1. Synchronization of papers across devices.

2. Notes on papers, that could also be shared.

3. A service to find publications of that paper (arXiv isn't always updated with this).

But you're probably better off discussing monetization with the other more intelligent experts here. I'm just a lurker :)