The document summarizes the system architecture and results of Vicomtech-ik4's system for lexical normalization of Spanish tweets. It achieved an accuracy of 65.15% on the test set and 65.42% on the development set, improvements of 5.09% and 4.42% respectively over the task baseline. Key aspects of the system included domain-adapted edit distances, a language model to rank candidate normalizations, and rule-based and dictionary resources to generate candidates, with the largest source of errors being gaps in domain dictionaries.