Good point. Normalization is deliberately scoped to 'what a human reads off the title page' rather than reconciling all possible metadata sources. LSH as a complementary fuzzy-matching layer for catalog reconciliation is exactly what the planned resolver at openusbn.org is designed to support: deterministic identifier as the anchor, probabilistic matching as the discovery tool.