I finally figure out how to create Word Wise klld database, turns out those unknown columns doesn't matter. I have updated the Proficiency and WordDumb code to use newly created klld files from Wiktionary data for Kindle books.
Add support of other gloss languages now seems more feasible. I found this cool project:
dbnary, they parses Wiktionary dump files in several languages and provides Turtle files which can be queried in RDF.