Survey - Related Work

From Stellar Deliverable 1.2

Co-author analysis and citation analysis is an important method when analyzing scientific communities. Ochoa et al. (2009) provides a very nice example of how such analysis can help provide greater insight into TEL research communities and collaborations, through visualizing and intuitively describing research community structure, focusing on TEL publications presented at recent EDMedia conferences. They focus on co-author analysis and visualization of these relations and provide interesting insights into collaboration networks in the TEL area. Wild et al (in press) used the same data corpus for a trend analysis in the EDMEDIA conference. By applying clustering techniques to the paper titles, they showed how certain technologies and approaches gained importance – including, among others, mobile learning, blended learning, portfolios, podcasts, game-based learning and assessment.

Similar introspective analyses have been applied to other research fields in the past. Henry et al (2007) provide an analysis of the area of human computer interaction, based on the four major HCI conferences, focusing on citation analysis that use data relating to these conferences (between conferences, articles and authors), word cloud visualizations to characterize the four conferences, and other visualizations that characterize collaboration and other networks. This paper does not rely on sophisticated mathematical network analysis modes but is a very good example of the power of visualization to make the structure of these networks explicit.

The approach we build upon in this paper, author co-citation analysis, has not yet been used widely despite its potential for detecting and clustering scientific communities based on the mathematical notion of factor analysis. One of the best papers and a good introduction to this approach is the paper by White et al, (White, H. D. and McCain, K. W. 1998). This study presents an extensive domain analysis of a discipline – information science – in terms of 120 top-cited authors, based on their papers from 1975 to 1995, with citations retrieved from Social Scisearch via DIALOG. Tables and graphics reveal the specialist nature of the discipline over 24 years, based on author co-citation analysis. The results show an interesting split of the field into two main specialties, which barely overlap, namely experimental retrieval/information retrieval and citation analysis. Included is also a dynamic analysis of the field, based on three 8-year-periods, which shows changes of authors and areas. The analysis is based on journal citations, but neglects important conferences such as the ACM SIGIR conference, the most relevant conference for the IR community. In contrast, the citation database used in our paper, CiteseerX, includes all important computer science conferences and workshops, providing a broad overview of computer science as it relates to TEL.

Using similar techniques, Chaomei Chen and Les Carr (1999) present an analysis of hypertext research based on the ACM Hypertext conference series, with papers included from 9 conferences over 10 years. About half of the citations in this series refer to papers from the same series, which points to a very homogeneous research community. Again, dynamic analysis using three time periods is included. Only citations within these conference series were considered, while we include citations from all conferences. Due to their restricted focus, the factors discovered represent a finely grained view of the hypertext research area (including subareas such as design models, hypertext writing, open hypermedia and information visualization), while our factors represent broader research communities, centered around one or a few community-centered conferences such as Adaptive Hypermedia or AIED.