kindageeky

Cloudstat is R in the cloud, the gallery app they have up so far inspects a twitter search, which reminds me of the clever presentation on twitter sentiment analysis with R by Jeffrey Breen.  To amuse yourself, click on the title link and search @united.  I hacked my way through the same presentation a few months back and at the time measured @united’s sentiment at 27%, maybe not surprising as airlines generally fare worse in approvals than the postal service.  Anyway, seems like there is a lot more brand management going on now as @united mentions itself in a positive light anytime a user registers for a contest for a trip to london.  Here is the topic cloud that cloudstat generated:

Yet when you look at the posts minus the brand management, you get a very different picture, for example this one …

@United’s plane was 3 hours late getting in to Boston. They asked me to fill out a survey, which is 404. http://t.co/lnpeEUwB

This whole space around sentiment analysis is very interesting, and especially hard to solve with posts as short as twitter. I have yet to see any evidence that sarcasm is being detected in sentiment analysis services, though there appears to be some promising research. 

Back to CloudStat, digging deeper there are various recipes (mostly against twitter) for doing analysis with R.  Apparently, I could work on my politeness.

> Result(x)

@kindageeky is ok polite, only better than 59 percent of twitter users. 

[1] "Ok."

> 

This is the output of CloudStat’s politeness tracker.  The notion of having an R language Cloud Robot is kinda cool, will have to look through the tutorials.