Visualizing Tone in Chat with IBM Watson and a Blink(1)

Re:Visualizing Tone in Chat with IBM Watson and a Blink(1) by k33n on Tuesday October 06, 2015 @04:39PM (#10342196)

You should use better examples. Just parsing text that says "I'm so sad", or "I'm happy!" isn't noteworthy.

[ Parent | on HN ]

Re:Visualizing Tone in Chat with IBM Watson and a Blink(1) by lotyrin on Tuesday October 06, 2015 @05:06PM (#10342395)

Unfortunately, Watson's sentiment detection is not capable of anything noteworthy.

[ Parent | on HN ]

Re:Visualizing Tone in Chat with IBM Watson and a Blink(1) by mtw on Tuesday October 06, 2015 @05:26PM (#10342525)

what would be a better sentiment analysis tool?

[ Parent | on HN ]

Re:Visualizing Tone in Chat with IBM Watson and a Blink(1) by nl on Tuesday October 06, 2015 @06:22PM (#10342836)

Metamind's sentiment analysis tool[1] is probably about the best you can get, especially for short text. Try the "Tweet" version (a tab on that page or available separately[2]).

Sentiment on short text is pretty hard. Metamind claims 81.73%[3], although that appears to be for general text (not just Tweets). Anything above ~70% accuracy would be competitive with the state of the art on Twitter data[4].

[1] https://www.metamind.io/classifiers/155

[2] https://www.metamind.io/language/twitter

[3] https://www.metamind.io/classifiers/155/stats

[4] http://www.lrec-conf.org/proceedings/lrec2014/pdf/483_Paper....

[ Parent | on HN ]