3

On occasion I'm in the wrong mood for writing e-mails, word documents and presentations. I know when that happens, and I "let it snow" a little before submitting. But if this was possible to detect, I'd love to have that feature. How far have we come in that area. Are we even close on detecting mood in language? And supply suggestion based on mood?

enter image description here

  • Given how much trouble the social media tracking industry is having with automated sentiment analysis of tweets and facebook updates, I don't think we've got the alogrithms to deliver this sort of functionality any time soon. – Racheet Feb 13 '14 at 17:27
  • 3
    The relevant keywords for google are "Automated sentiment analysis" and "Natural Language Processing" – Racheet Feb 13 '14 at 17:27
2

Short answer:

Yes it is possible to detect. Spam filters can classify spam material and it is used in data mining industries.

No, it is no way near ready for use for alerting when writing.

Long answer:

As Racheet said, this is an aspect of "Automated sentiment analysis" and "Natural Language Processing". It is possible to detect passive aggressive language in text. The problem is that such a system needs training. The system needs to know how you use language and needs to understand the context.

An example is the phrase "let it pour". A farmer may say "let it pour" in an email without referring to farming directly. This is a hopeful statement. A disgruntled worker may say "let it pour" in an email which could be taken as very aggressive. Additionally, This phrase may not have the same meanings in all countries.

If the emails from the farmer and disgruntled worker both use the phrase "let it pour" without stating they are a farmer or disgruntled worker, there is no context. There would need to be prior written text for the computer (or human for that matter) to work off.

This is not 100% accurate and the type of people using text editing programs expect the suggestions to be correct.

Additionally it would need to be trained for each new person. This training would involve the user stating if examples were passive aggressive or not. This would require 100+ examples of text. This is also hard for end users.

1

Are we even close on detecting mood in language?

No, we're not close to automatically detecting things as subtle and nuanced as mood. It simply requires a much greater understanding of language use than we're able to encode now. We're just starting to get to the point of machine understanding of clear unambiguous human language.

Another problem is that the expressions of mood varies so much between cultures and dialects. A mood analyzer would probably require significant training by each individual user.

Natural language processing is the next great frontier, and tons of money and talent is being applied to it, but don't expect a machine to understand nuanced things like sarcasm, satire, double-entendre, etc. anytime soon.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.