Yes but parsing theorems (at least at the syntactic level) should be no harder than parsing English, and we do lots of reasonably smart text mining, and even some reasonable natural language translation.
Math theorems are hard-going for many humans. Machines think differently - but may well find them challenging too. I'm not sure this area is particularly low-hanging.
Does anyone know of work that attempts to build a theorem prover by learning-from-examples? I'm imagining extracting a large corpus of theorems from back issues of mathematical journals, then applying unsupervised structure discovery techniques from machine learning to discover recurring patterns.
Perhaps a model of the "set of theorems that humans tend to produce" would be helpful in proving new theorems.
The unsupervised-structure-discovery bit does seem within the realm of current machine learning.
Any references to related work?