3 ms·

There are companies that provide text analysis services to determine the 'temperature' of the text or mood of the writer. The x axis of the graph is unexplaine

There are companies that provide text analysis services to determine the 'temperature' of the text or mood of the writer. The x axis of the graph is unexplained, but the idea (of being able to measure sentiment) isn't crazy.

sogrady 15y ago
it's actually just the code from here: http://jeffreybreen.wordpress.com/2011/07/04/twitter-text-mining-r-slides http://jeffreybreen.wordpress.com/2011/07/04/twitter-text-mi... applied against a mongodb query. obviously not statistically representative; but rather an indication of what an acceptable profile looks like.
- mechanical_fish 15y ago
  Of course you can measure customer sentiment. You can even measure the correlation between your brand name and a selection of happy words as they appear on Twitter. What you should probably not do, however, is elide the essential difference between these two measurements. Twitter is not a representative sample of anything but Twitter. Much of Twitter is spammers and shills, not customers. A lot of Tweets aren't even from humans. And people can mention MongoDB without knowing the slightest thing about it, and I am sure many do. I'm doing it now.
  - sogrady 15y ago
    where we agree then, i think, is the fact that Twitter's just another datapoint. and from a statistical perspective, it cannot be considered representative of the wider population for any number of reasons, not least the fact that it's observational rather than a random sample. that said, it's an interesting proxy for examining questions of sentiment, and its predictive ability has been examined several times academically within specific contexts (elections and markets - e.g. this one by Cornell http://arxiv.org/abs/1010.3003 http://arxiv.org/abs/1010.3003) and found to be relatively accurate. so agreed, it's not representative. but that doesn't mean it's not interesting and potentially useful.

sogrady 15y ago

it's actually just the code from here: http://jeffreybreen.wordpress.com/2011/07/04/twitter-text-mining-r-slides http://jeffreybreen.wordpress.com/2011/07/04/twitter-text-mi... applied against a mongodb query. obviously not statistically representative; but rather an indication of what an acceptable profile looks like.

mechanical_fish 15y ago
Of course you can measure customer sentiment. You can even measure the correlation between your brand name and a selection of happy words as they appear on Twitter. What you should probably not do, however, is elide the essential difference between these two measurements. Twitter is not a representative sample of anything but Twitter. Much of Twitter is spammers and shills, not customers. A lot of Tweets aren't even from humans. And people can mention MongoDB without knowing the slightest thing about it, and I am sure many do. I'm doing it now.
- sogrady 15y ago
  where we agree then, i think, is the fact that Twitter's just another datapoint. and from a statistical perspective, it cannot be considered representative of the wider population for any number of reasons, not least the fact that it's observational rather than a random sample. that said, it's an interesting proxy for examining questions of sentiment, and its predictive ability has been examined several times academically within specific contexts (elections and markets - e.g. this one by Cornell http://arxiv.org/abs/1010.3003 http://arxiv.org/abs/1010.3003) and found to be relatively accurate. so agreed, it's not representative. but that doesn't mean it's not interesting and potentially useful.