I noticed that Lingua identifies wrongly a text which includes portion of foreign words

This is an example of a Korean text, which includes the string "CA" which is not Korean (probably this represents the initials of a person)
( 웃음 ) CA : 실패하는군요 . 안타깝네요 .

This text is identified as "Romanian".
This is a bit strange since there are 3 Koreans tokens and only one (probably) not Korean.

Actually, also the punctuation marks should be "Korean".

Any idea why this wrong identification occurs?