Reader Comments

Post a new comment on this article

So what?

Posted by ronewolf on 25 Jan 2011 at 17:42 GMT

I have little problem with your methodology, indeed what you did regarding automated examination of tweets might be of interest in computational and machine-analysis circles. However I don't see why your report is of any interest regarding medicine. For one thing, similar observations were made years ago regarding the use of Google trends to detect public interest in medical events. Yet, you don't reference these studies. Worse, your study lacks geo-information (which Google Trends provides) and so is even closer to useless. People also generate huge tweet volumes regarding Ashton Kucher, so what?

I did find one of the graphs to be of interest. Figure 4 - Scatterplot of tweets sharing personal experiences and USA H1N1 incidence rate.- as it seems to further back up my critique. The graph shows that out of the 2M tweets recorded, only a few tens of thousands were identified as related to personal experience with H1N1. That is not a high ratio ( a few %). Additionally of note, the actual incidence of H1N1 was about half the # of tweets. Indeed the virus was much more of a scare than it was an epidemic. In that regard, what did we learn from social media? Only that peeps love to relay reports that the sky is falling.... Some things never change....

No competing interests declared.

RE: So what?

geysenba replied to ronewolf on 25 Jan 2011 at 21:45 GMT

In response to ronewolf:
1. What is the "interest regarding medicine": See - Quote: "When public health officials talk about social media and using social media, what they mean is basically getting our messages out to the public but that's not what the social web is all about. It is only half the equation. The other half is also using this to learn from what people are talking about on the social web and using that information to customize and tailor your own public health messages."
Out intention is to develop tool/methods allowing public health professionals to do that.

2. "using Google trends to detect public interest not referenced": Not sure what particular study you are referring to, but for all I know ref 6 (my own study, initiated in 2003 and published in 2006) was the first study suggesting to use Google searches for public health purposes (this was before Google trends became available, and many years before Google Flutrends).
Further references are in the bibliography of ref 7.
To our knowledge, this is the first study using tweets. Search data are "sparse" data, whereas tweets are much richer and allow a distinction of the users' intention, behavior, knowledge, attitude, attention to topics.

3. No geocoding: Although not done in this study, geocoding is possible and we are using it in other projects.

No competing interests declared.

RE: RE: So what?

hans_jehann replied to geysenba on 29 Sep 2014 at 10:26 GMT

Looking forward to see other projects that incorporate geocoding, I believe it's a very important point when it comes to studies that take place on Google.It's not as easy to do with Twitter as a recent study showed most users do not add a real location in their profile, and location must be inferred from other factors.

Hans Jehann

No competing interests declared.