In this lightning talk I will present a counter-intuitive use of reading normalization to improve the results of automated collation.