We discuss the design, building and evaluation of a method to access theinformation of a person, using his name as a search key, even if it has deformations. We present a similarity function, the DEA function, based on the probabilities of the edit operations accordingly to the involved letters and their position, and using a variable threshold. The efficacy of DEA is quantitatively evaluated, without human relevance judgments, very superior to the efficacy of known methods. A very efficient approximate search technique for the DEA function is also presented based on a compacted trie-tree structure.