Detect unreadable unicode text
Ever encounter any text when crawl text from the internet or inside your raw corpus?
srtytyrtyrty
Á¶À̽ÃƼ, ¡®3on3 ÇÁ¸®½ºÅ¸ÀÏ¡¯ 2Á¾ÀÇ ¿¡µð¼Ç ¹øµé Ãâ½Ã
��>+ٽT}$@�������Э�����ٗ_���=���e��
This is what flawunicode aims to pick these out for you. flawunicode ranks each unicode text and output a score of -1 to 1 which indicates the "completeness" of the unicode text. If the text has a score of lower than 0.4, it is likely this text is not readable by human.
import flawunicode
text = "fdsfdxvdhjkf"
flawunicode.detect(text)
>> 0.2727272727272727
flawunicode.detect("Hello World!")
>> 0.6439393939393939The underlying statistic came from news corpus in currents api database. So social network style text maybe rank with low score. You just need to calculate your own frequently used bi-gram characters and it should be fine.