edited to add: Hello new visitors! Eek! I had no idea there would be so much interest. Please be advised: Twitter comments and text mining are hardly exact sciences, and this was an exercise to see whether we could get a simple algorithm to detect positive and negative commentary.
The takeaway is that people complain about transit rather a lot–particularly when it’s 100+ degrees out!
IOW, these are NOT rankings of transit system performance! They are rankings of how much, on average, Twitterers complain/compliment the services or individuals examined via the feeds.
I haven’t analyzed the word clusters yet, but it’s pretty clear: one of the most common things to complain about: other passengers! (not related to transit company performance!)
So with those caveats:
I recently scraped some Twitter feeds to gauge the sentiment expressed about three control groups and two groups of mobility providers: airlines and transit providers. The controls: celebrities, politicians, and “villains”.
Next up: what are people saying as they kvetch about transit?
Schweitzer, L. 2011. “How are we doing? Transit and Airline Customer Sentiment Expressed Via Twitter”. Working paper. University of Southern California School of Policy, Planning and Development.
Impressive. I’m eager to see more / read the article.
(I assume you already thought of this) I’m curious what type of things people tweet’ed about. I can imagine some of them would be be useful for transit agencies and/or transit advocates to monitor late vehicles, unclean equipment, crime/safety, customer service, responses to new service/technology/equipment, and so on. Did you categorize them or do you plan to do so?
I don’t categorize them in this paper, but I will do so in subsequent work.
Dr. Schweitzer, I’m the editor of a website and magazine for public transit riders in Portland, and I’d love to reprint this image with credit, link to your site and description of your research. Would this be all right?
Sure, but I should probably talk about the research a bit more–this is only a few scrapes of the Twitter from March, June, and July, and it’s about Twitter commentary–hardly scientific. But you can link to it and chat about it if you like, as long are you are careful to caveat.
Makes sense. It’s up! Thank you. http://portlandafoot.org/2011/08/trimet-americas-most-popular-transit-agency/
What search strings did you use for DC’s Metro?
dcmetro @metroopensdoors metroopensdoors
You should try #wmata, @wmata, @unsuckdcmetro, and @fixwmata.
did you use dcmetro for searching for metro, or did you use wmata? Not sure which is used more often, but wmata is the actual name, and is used frequently.
I did use wmata, but it didn’t yield many comments, so I didn’t fret about following up on that usage. dcmetro tag yielded enough to fill sample (I scraped 500 tweets at a time–dcmetro gets that consistently).
You’re hitting the nail on the head–it’s tough to get a good search. put together, and sometimes people flag both #dcmetro and #wmata in the same comment, and two scrapes with grab both comments, and so you have to dig through for duplicates if you do that.