Learning Lifted Action Models from Traces with Minimal Information About Actions and States