In the English Wiktionary dump, I have seen a lot of non-alphabet entries. I think we can filter out these entries to reduce the workload…
