- by magneticnorth0, Data Doctor
- 3/22/2014 10:13:51 AM
@Beth Like we marketers always say, it depends on the market. If your market is the type to express themselves on Twitter, the study should be done there; if not, then we'll go where the market is more expressive of their emotions. For the sake of ease, though, I think analysis that relies on publicly-available data will be more ideal.
- by BethSchultz, Blogger
- 3/14/2014 1:58:59 PM
@magneticnorth, thanks for sharing your perspective as a marketer. So when you think about sentiment analysis and the value it could provide in helping marketers evaluate and execute branding campaigns on profitability, are you thinking in terms of what's being said on social networks or in survey verbatim responses, or in focus groups, call center records, and so on? In other words, what do you consider the best source to mine for customer sentiment?
- by magneticnorth0, Data Doctor
- 3/13/2014 8:09:05 PM
Marketers might pay a pretty penny to be able to forecast with high reliability how a few well-placed sentiments might spread!
As a marketer, Beth, I can tell you that many of my colleagues will happily put this type of research in their budget (assuming it's fairly reliable, of course). But personally, the value I see in this technology is primarily in its ability to quantify a number of variables involved in branding. I remember a professor of mine saying that "perceived value" is the only brand attribute that's been proven to contribute to sales. There are many others—e.g. brand relationship, emotional loyalty—that we intuitively know are contributors to the bottom line, but they're just too hard to measure. I think sentiment analysis will finally allow marketers to execute and evaluate branding campaigns based on profitability, whether in the short or long run.
- 3/7/2014 11:24:03 AM
That seems to be the standard MO for politicians everywhere, Seth. Sometimes the diversion of attention and topic is subtle, but often it's appallingly blatant. Software like that of Beyond Verbal or Emotient, though, should be able to detect this kind of obfuscation.
- by SethBreedlove, Data Doctor
- 3/6/2014 6:09:22 PM
@ Beth, politicians pay good money for speech writers. It is in the interviews when they are caught off guard that the truth comes out. Especially when they refuse to answer a question, but keep smiling.
I was once told by a reporter was that it wasn't important what question you were asked but rather what answer you gave. That way someone flipping through the channel only hears what you wanted them to hear.
- 3/6/2014 5:17:38 PM
Fascinating analyses of Super Bowl ads and temporal reactions to different portions of ads.
"It enables a fine-grained temporal analysis that would be much harder to do with human annotators."
- 3/6/2014 5:07:03 PM
Emotient - measuring emotions via facial expression:
Microexpressions - joy - anger - disgust - contempt (snort) -
Engagement level of a student during a class - high correlation with test results.
- 3/6/2014 4:56:33 PM
Let machines understand our emotions based on our vocal intonations, and get a better understanding of our own selves. Tutoring, dating, TV, product recommendations, you name it.