Multi-domain knowledge reasoning