The last time I wrote something, I was in Hong Kong and just wanted to get a post out quickly for everyone to see. Now, today, I’m home and I can fill in the gaps between the last blog written in China, and now.
The Web Science Winter School
The Winter School was all about working with other students from Singapore, Korea and Tsinghua University. We were expected to work with data, provided according to four themes, and produce some form of interactive model using the data. The themes were:
- Innovation on the web;
- Citizen engagement via the web;
- Natural disaster management; and
- Indian elections.
As a group, we decided to make sure there was at least one Southampton student in each group. I decided to go with group 4, as their focus was on sentiment analysis as a method for both measuring the positivity or otherwise of the tweets through the campaign (the data set was drawn from tweets made in the six months before the election data, and a couple of weeks after) and to look at creating a model that could be used to predict the outcome.
Each group had a tutor appointed to provide guidance. Ours was Professor Nie Liqiang from the National University of Singapore. Oh dear. He was very clear about what he wanted us to do, and summarised it all on the board for us. And then left. Apparently, he brought his wife on the trip with him. This fact may or may not be connected with the lack of attention we received as a group. When he did appear, it was to give us more ‘directions’. Naturally, his own students did as they were asked and, although they were lovely to work with, I have to say that they did exactly the same thing with the data as has been done many times before. Which is exactly what the Professor wanted. What we were meant to do was discuss ways in which we could present and utilise the data to reveal something different, but he wasn’t having any of it. I did a bit of research on my own, and found something that we could have developed, but all I could argue for was that it was included in the ‘future work’ slide of our presentation. And I argued as hard as I could without causing a diplomatic incident.
If I tell you he referred to me as ‘lady’ and ‘sister’ when he wanted to speak to me, that will probably give you some idea of what I was up against.
I smiled in the face of provocation, and remembered this was all about collaborative work and positive relationships. The other groups did much better though.
Group 1 focused on patents as a way of measuring innovation, and developed an interactive model where, put simply, the number of patents applied for / granted together with the total spending on research and development by countries could be selected to provide an innovation ‘score’ (underneath the user interface, the programme is using data sets drawn ‘live’ from data published on the www).
Group 2 decided to use data gathered from the Chinese versions of sites covering the cost of renting property, buying goods etc. to calculate the cost of living of a region / city / neighbourhood and provide the output as an overall score. Again, this was presented as an interactive, working model.
Group 3 were tasking with looking at the fires in Indonesia and measuring their impact on Singapore. They used a set of tweets from Twitter to create a model that would measure the degree of sentiment expressed in relevant tweets, and display the degree of pollution in different areas.
All of these were presented to a panel of academics on Friday. The winning group? Group 2.
As I said, the group I was working with only focused on processing the data we’d been given in ways that have already been done many times. What would have been more interesting would have been if we could have built a working model that extracted a sample of tweets and provided a real-time analysis of the conversations going on around the selected candidates in terms of topic and sentiment. Following the results of the UK general election, one of the articles I found from The Guardian newspaper suggested that “had the forecasts been different, then the nightly news bulletins would surely have concentrated rather more on the vast spending cuts to come…” and “…had the Tories been confident of winning outright, they might have written a different manifesto…” (which in no way excuses them for lazy journalism). We were able to model the changes in the sentiment of conversations based around topics over the course of six or seven months – more work at refining the data set and topic modelling might have produced a more nuanced and interesting result that could have fed The Guardian the information it claims it was looking for.
It does, though, raise the issue of how influence is propogated across networks such as those that exist in twitter. The traditional flow of information and influence from the news media and other ‘popular’ authorities to ordinary citizens may now be permanently disrupted by social networks and other forms of engagement such as comments left under online news stories. The flow may now be more like a tennis match taking place on the court of the www: the ball representing topics of conversation that are hit backwards and forwards from citizen to media organisations and back again. During the game, what change takes place in the ball, the players and/or the court?
More to follow…..