The corpus i found -even thought creating a KB didn’t work out- was from extracting wikipedia with a script.
This might work for you.