Research in digital preservation and machine learning

Hello guys,

I had a talk with my prof that day about digital preservation. It seems to be one of the big topics for the next years and I am extremely interested to write about a problem which goes with this topic. At the moment I am missing some interesting ideas, which connect to big data or machine learning. Therefore my question:

What are for your the interesting problems in digital preservation in connection to machine learning?

Would appreciate ideas/papers/keywords!