We met, we tweeted, we archived... then what?
We're all getting increasingly used to using Twitter as a back-channel at events. Indeed, it is now relatively uncommon to turn up for an event at which there isn't both a pre-announced hashtag and an active circle of twitterers already in attendance.
We also recognise that Twitter doesn't leave our tweets lying around for very long in the Twitter search engine and that if we want some kind of a more persistent and accessible record of Twitter activity at an event then we need to arrange for a copy of all the tweets to be archived somewhere. Normally, in my experience at least, TwapperKeeper is currently used to create that archive.
So far, so good... but then what? Offering a vanilla view of a few thousand tweets is potentially useful for those who want to delve into the detail, but it hardly provides an easy to grasp summary of the event. How can we present a view of the Twitter archive such that a summary is offered without the need to read every tweet?
There are some obvious simple things that can be done with the RSS feed of tweets offered by TwapperKeeper, and I've knocked together a quick demonstrator to show the possibilities...
Firstly, we can count up the total number of tweets, twitterers, hashtags and URLs tweeted during the event. That gives us an overall feel for how 'significant' the use of Twitter was.
Secondly, we can display a list of the people who tweeted and were @replied the most (in Twitter parlance, an @reply is a tweet that directly mentions another Twitter user). We can also see who was involved in most 'conversations' (exchanges of @replies between any two Twitter users). That gives us a feel for who was tweeting the 'loudest'.
Thirdly, we can look at what hashtags and URLs were tweeted the most. That gives us a feel for the topics and resources most related to the topic of the event.
And finally, we can unpick the individual words used in the Twitter archive, providing a kind of 'word cloud' for the event.
None of which is rocket science... but it is potentially useful nonetheless. Here are such summaries for the Repositories and the Cloud meeting that we recently organised with the JISC, for the JISC Dev8D event, and for the National Digital Inclusion 2010 conference (based on the associated TwapperKeeper archives for each of the events).
In a follow-up post to the NDI10 event, After the event, and a subsequent message to the UK Government Data Developers Google Group, Alex Coley suggests going further:
I wondered if a flash based tool could be used to map sentiment by session/topic by giving positive/negative meanings to words and applying this to tweet traffic. Perhaps some real meaning and value could come out of conferences that anyone can access and use.
Sounds interesting, though I have no idea how to implement it!
Dave Challis of the Southampton ECS Web Team has also written up a couple of blog posts following Dev8D, A first look at the dev8d twitter network and Dev8D twitter network, part 2, in which he discusses the analysis of Twitter to see how people's social networks evolve during an event. Fascinating stuff!