Nicolas Marie and I will be presenting at the WebSci'12 poster session tonight. The poster is based on our note, written with Evangelos Kalampokis, entitled Dissecting the Butterfly: Representation of Disciplines Publishing at the Web Science Conference Series. In a nutshell, we're presenting the method we used to gain initial insight into disciplinary representation at the Web Science conference series, along with initial results from applying that method. It's not uncommon for people in and around the Web Science arena to have conversations about what we mean by 'Web Science', and about disciplinary diversity and representation -- asking questions about what disciplines are more or less present, whether one discipline is drowning others out, and so on. We did some empirical work, grounding that discussion in data. As well as loaning some credence to the discussions of diversity, it helps us:
- better communicate within our community
- reach out to other communities with which we want to engage
- identify problems with disciplinary representation -- see what types of research are missing and what collaborations we might seek
We looked at WebSci papers from 2009 to 2011, conducting topic extraction and then graphing links (papers) between those topics. We identified clusters of related topics, and calculated betweenness centrality (showing which topics have a high likelihood of bridging disciplines -- that is, which topics are key areas of collaboration). Finally, we mapped our findings to the Web Science Butterfly, the diagram showing disciplines that are purportedly part of Web Science. The diagram was used early on to convey the WebSci vision, and is sometimes used to describe the discipline -- we wanted to examine its accuracy. The heat map below shows what topics are more and less represented in the Web Science community (mapped onto the diagram used early on to convey the WebSci vision). Our paper is focused on the methods rather than this specific outcome -- practical constraints meant the final analysis included only 69 of the 91 papers in the corpus, and we need a stronger method for determining whether a discipline is strongly or weakly present in the community.
I'm really happy about this piece of work. To our knowledge, it's the first exploration of disciplinary representation within Web Science and these early results provide a good first insight. Furthermore, we're contributing a method that others can pick up and use. There are some important questions we should be asking in this area -- for instance, we should be checking out the impact of collocating with specific conferences, whether there are disciplinary differences between posters and papers, and indeed which disciplines collaborate. Swing by our poster tonight to learn more!