Looks like NAACL is up on ACL Anthology, and so too is our ( Junyoung Lee + @sxm.bsky.social + Naoaki Okazaki) paper on Jamo-level Subword Tokenization for Korean Machine Translation (from LoResMT).

#tokenization #korean #nlp

Vivaldi Mail really is intolerant to fat fingers - wonder if I should file a feature request for a manual "send all in draft" button...

Been sifting through common crawl a bit due to a side project. Fresh reminder that the internet is full of spam and garbage...

I'm fairly confident I can get any model that has seen common crawl to write online casino and viagra spam in Korean

Starting to post here as Threads made a conscious decision to not have a web client. Let's see which one sticks. (I like the concept of Mastodon, but the clients frankly are sort of crap.)