There are too many word "alamy" in Union14M-U in cc_lmdb

[zhangyushao0]

Observation:
I have checked randomly about 1000 images of cc_lmdb, it turns out that about 20% of cc_lmdb is word "alamy".

Questions/Concerns:

1. Is this expected behavior for cc_lmdb?
2. Could this skew model training?