5 ms·

Computer program uses Twitter to 'map mood of nation'

20 points by eplanit 13y ago

TomGullen 13y ago
Not map mood of nation, map mood of Twitter users in a nation. Big difference.
- diminoten 13y ago
  It's a large overlap.
  - JoeAltmaier 13y ago
    Its a tiny, tiny overlap. Check this out for twitter demographics: http://socialmediatoday.com/index.php?q=SMC/78505 http://socialmediatoday.com/index.php?q=SMC/78505
    - diminoten 13y ago
      20% is hardly tiny.
      - JoeAltmaier 13y ago
        Read carefully. 20% of 20% == 4%
msy 13y ago
As a human, out of curiosity I've previously tried to classify random tweets into emotions and failed, there just isn't enough context to make the judgements not arbitrary. In consequence I'm deeply suspicious of software that claims to be able to do so.
- Shank 13y ago
  I worked with a lot of Twitter data over the summer as part of an internship. I can say that signal to noise ratio is so abysmally inverted, this is almost certainly doing a lot of filtering just to get meaningful source data. For instance, you have to filter out most posts with links (because they tend to have very little associated data, or are spam), anything in languages you don't have parsers or datasets for, etc. Then you can cut on character count (because there are a surprising amount of useless tweets that are under 5-9 characters)... As others have and will continue to state, this further cuts down the sample size of an already fairly small sample (500+ million in the world[0]). If you want to be specific and only look at a single nation, you're cutting that sample size based on who has location turned on and/or broadcasts it in their profile. This is a significantly smaller number. Edit: This is obviously a better sample size than a random poll of individuals, however, but that doesn't mean that the cumulative quality will be drastically higher. [0] - http://www.statisticbrain.com/twitter-statistics/ http://www.statisticbrain.com/twitter-statistics/ (sources from three different input points)
danso 13y ago
Statistics 101...sheesh. Not only is it limited to users of Twitter (and not the nation), but it's limited to users who geocode their tweets, which may be as little 1 to 2% of them. After the presidential election, Gallup overhauled its polling methodology. Why? Because they significantly overestimated Mitt Romney's popularity, and one of the factors may have been relying too much on landline polling, as older (i.e. more conservative) people are disproportionately likely to still use landlines. If Gallup's predictions are as sensitive to a factor like that, I imagine this neophyte tech/science company is going to be way off in their measurements.
troebr 13y ago
It's the first exercise in datasci-001 on coursera... Not much of an invention. https://class.coursera.org/datasci-001 https://class.coursera.org/datasci-001
primaryobjects 13y ago
This is a timely article, as my current project deals with Twitter sentiment analysis http://www.sentimentview.com http://www.sentimentview.com. Although, I'm not classifying by specific emotion type as the article describes. In all, I'm seeing accuracy rates of around 80%. This is in comparison to a brute-force word list, which scored an accuracy of 62%. I'd be curious to see what kind of accuracy the researches in the article are achieving, with so many sub-division classifications. Not to mention, how did they derive their initial model?
- nhebb 13y ago
  How do you detect sarcasm?
tomrod 13y ago
I love this. As an economist, I'm very interested on the correlation between this distribution of rhetorical reactions Twitter captures and economic happenings. I first realized there might be a connection with the BP oil spill. One would expect that oil company and possibly general energy stocks would fall after such a negative happening. Can we predict the impact of such circumstances in the Tokyo exchange too, based on the rhetoric we can measure in each nation (US and Japan)? Oh man, the future is exciting as big data collection and analysis mature.
lukeh 13y ago
What about We Feel Fine [1], saw a TED talk on this years ago and loved it. [1] http://www.wefeelfine.org http://www.wefeelfine.org

20 points by eplanit 13y ago

TomGullen 13y ago

Not map mood of nation, map mood of Twitter users in a nation. Big difference.

diminoten 13y ago
It's a large overlap.
- JoeAltmaier 13y ago
  Its a tiny, tiny overlap. Check this out for twitter demographics: http://socialmediatoday.com/index.php?q=SMC/78505 http://socialmediatoday.com/index.php?q=SMC/78505
  - diminoten 13y ago
    20% is hardly tiny.
    - JoeAltmaier 13y ago
      Read carefully. 20% of 20% == 4%

msy 13y ago

As a human, out of curiosity I've previously tried to classify random tweets into emotions and failed, there just isn't enough context to make the judgements not arbitrary. In consequence I'm deeply suspicious of software that claims to be able to do so.

Shank 13y ago
I worked with a lot of Twitter data over the summer as part of an internship. I can say that signal to noise ratio is so abysmally inverted, this is almost certainly doing a lot of filtering just to get meaningful source data. For instance, you have to filter out most posts with links (because they tend to have very little associated data, or are spam), anything in languages you don't have parsers or datasets for, etc. Then you can cut on character count (because there are a surprising amount of useless tweets that are under 5-9 characters)... As others have and will continue to state, this further cuts down the sample size of an already fairly small sample (500+ million in the world[0]). If you want to be specific and only look at a single nation, you're cutting that sample size based on who has location turned on and/or broadcasts it in their profile. This is a significantly smaller number. Edit: This is obviously a better sample size than a random poll of individuals, however, but that doesn't mean that the cumulative quality will be drastically higher. [0] - http://www.statisticbrain.com/twitter-statistics/ http://www.statisticbrain.com/twitter-statistics/ (sources from three different input points)

danso 13y ago

Statistics 101...sheesh. Not only is it limited to users of Twitter (and not the nation), but it's limited to users who geocode their tweets, which may be as little 1 to 2% of them. After the presidential election, Gallup overhauled its polling methodology. Why? Because they significantly overestimated Mitt Romney's popularity, and one of the factors may have been relying too much on landline polling, as older (i.e. more conservative) people are disproportionately likely to still use landlines. If Gallup's predictions are as sensitive to a factor like that, I imagine this neophyte tech/science company is going to be way off in their measurements.

troebr 13y ago

It's the first exercise in datasci-001 on coursera... Not much of an invention. https://class.coursera.org/datasci-001 https://class.coursera.org/datasci-001

primaryobjects 13y ago

This is a timely article, as my current project deals with Twitter sentiment analysis http://www.sentimentview.com http://www.sentimentview.com. Although, I'm not classifying by specific emotion type as the article describes. In all, I'm seeing accuracy rates of around 80%. This is in comparison to a brute-force word list, which scored an accuracy of 62%. I'd be curious to see what kind of accuracy the researches in the article are achieving, with so many sub-division classifications. Not to mention, how did they derive their initial model?

nhebb 13y ago
How do you detect sarcasm?

tomrod 13y ago

I love this. As an economist, I'm very interested on the correlation between this distribution of rhetorical reactions Twitter captures and economic happenings. I first realized there might be a connection with the BP oil spill. One would expect that oil company and possibly general energy stocks would fall after such a negative happening. Can we predict the impact of such circumstances in the Tokyo exchange too, based on the rhetoric we can measure in each nation (US and Japan)? Oh man, the future is exciting as big data collection and analysis mature.

lukeh 13y ago

What about We Feel Fine [1], saw a TED talk on this years ago and loved it. [1] http://www.wefeelfine.org http://www.wefeelfine.org