I’m assuming that you want to find data where the streetname might be slightly misspelled? E.g. it might be “CALE FCO POLANCO” or “CALL FCO POLANCO”?
What you probably need for this kind of query is to use Full Text Search, which is more language aware and can handle fuzziness. If you decide to try LIKE, you’re going to run into a whole bunch of edge cases.
There is no N1QL function for levenshtein distance (yet) as far as I know. A full text search can do fuzzy matching and exact matching.
Also, if you want to use both together, as of Couchbase Server 6.5 (currently in beta), you can call Full Text Search (FTS) indexes from N1QL queries.