Solved

Emoji ♥️😂

  • 20 May 2021
  • 3 replies
  • 58 views

Userlevel 1
Badge
  • New Participant
  • 3 replies

What is the expected behavior around emoji’s within Document analysis and Document classification?

icon

Best answer by bmunz 20 May 2021, 20:15

View original

This topic has been closed for comments

3 replies

Userlevel 4
Badge +2

As of now, the emoji (if it’s UTF-8) will be ingested into the engine with the other text.  Unfortunately, the emoji will then lose its meaning and be converted into an innocuous character and thus not be analyzed as text.  This is something we’re looking to update on the product side in the future.  I suppose a workaround could be to convert the emoji into representative text before running it through the NL API?  That way the sentiment could be captured.

Userlevel 1
Badge

@bmunz Thanks for the reply!  Now when you say “convert the emoji into representative text”, do you mean “CLDR Short Name”?

Userlevel 4
Badge +2

To be clear this is just an idea that I haven’t tested, but yeah I think if the emoji was converted into text like “smiling face” or something similar, it could help the NL API understand the meaning represented by the emoji to better analyze the text.  The emotional traits taxonomy will identify emoticons, I think, but the others don’t YET so it might be best to try to use text for now.

I talked to R&D about this and they said that they’re going to work to get emojis properly identified in the API ASAP.  Stay tuned!