Monday, October 19, 2020

Censoring repugnant words by algorithm

 Some people like to say things that other people think they shouldn't say.  In the age of the internet, politeness can be (somewhat) automated, by banning certain words.  But of course, words have contexts. Here's a funny story from the Guardian:

Overzealous profanity filter bans paleontologists from talking about bones--A virtual conference was thrown into confusion when the platform hosting the event came with a pre-packaged ‘naughty word’ censor by Poppy Noor.

"Participants in a virtual paleontology session found themselves caught between a rock and a hard place last week, when a profanity filter prevented them from using certain words – such as bone, pubic, stream and, er, beaver – during an online conference.

"The US-based Society of Vertebrate Paleontology (SVP) held its annual meeting virtually this year due to the pandemic, but soon found its audience stifled when they tried to use particular words.

"Convey Services, which was was handling the conference, used a “naughty-word filter,” for the conference, outlawing a pre-selected list of words.

"“Words like ‘bone’, ‘pubic’, and ‘stream’ are frankly ridiculous to ban in a field where we regularly find pubic bones in streams,” said Brigid Christison, a master’s student in biology attending the event


"Some discovered bias in the algorithm, too. Jack Tseng, a vertebrate paleontologist from the University of Berkley pointed out that the filter had banned the common surname Wang but not Johnson – even though both are frequently used as slang words to describe a man’s genitals."


Here's Dr. Tseng's tweet:

Z. Jack Tseng, @Tseng_ZJ

"Wang" is banned but not "Johnson" (both used as slangs). This western-centric filter erasing the surname of 90+ million Chinese but not <2 million people of European descent is unexpectedly on brand for 2020,  ! My PhD advisor is X. **** by the way. "


Previous related posts:

HT: Muriel Niederle

No comments: