Topic Modeling of Research Fields: An Interdisciplinary Perspective

International Conference RANLP 2009 - Borovets, Bulgaria, pages 337–342 Topic Modeling of Research Fields: An Interdisciplinary Perspective Michael Paul and Roxana Girju University of Illinois at Urbana-Champaign {mjpaul2,girju}@illinois.edu Abstract This paper addresses the problem of scientific research analysis. We use the topic model Latent Dirichlet Allocation [2] and a novel classifier to classify re- search papers based on topic and language. Moreover, we show various insightful statistics and correlations within and across three research fields: Linguistics, Computational Linguistics, and Education. In particu- lar, we show how topics change over time within each field, what relations and influences exist between top- ics within and across fields, as well as what trends can be established for some of the world’s natural lan- guages. Finally, we talk about trend prediction and topic suggestion as future extensions of this research. Keywords topic models; scientific research analysis; statistical approaches 1 Introduction No one can predict (at least not in detail) the stringent issues that science and society will consider in the next decades. However, if we look at some top-priority issues of today - such as health, economy, homeland security, stem-cell research, science teaching - and pressing research questions, such as how to enhance child development and learning and even how to make sense of the huge amount of information with which we deal daily, we can say that future topics will be so complex as to require insights from multiple disciplines. Interdisciplinary research will thus facilitate the integra- tion of information, data, techniques, tools, perspectives, concepts, and/or theories from two or more disciplines or sources of specialized knowledge to advance fundamental understanding or to solve problems whose solutions are be- yond the scope of a single discipline or field of research. We believe that like other disciplines, Computational Linguistics (CL) will drastically benefit from an interdisci- plinary perspective. This research is part of a larger project whose goal is to design a system which will help foster interdisciplinary research in order to make breakthrough predictions for future directions. The system is also in- tended to promote interdisciplinary collaborations by pro- viding novel topic suggestions to professionals who would like to engage in research discussions with other parties, but who are not familiar with those areas. This paper presents details about the current version of our system which researches a set of three fields: Linguis- tics, Computational Linguistics, and Education (we include here Educational psychology). Based on various topic models which classify research papers into topic and lan- guage categories, the system displays a series of statistics, correlations, and graphics which show the dynamics of top- ics and local and global trends based on the proceedings of the major conferences and journals in the three fields over many years. In particular, we show how topics change over time within each field, what relations exist between topics, what temporal correlations and topic influences can be determined across fields, as well as what trends can be established for some of the world’s natural languages. Fi- nally, we …

Content truncated to 500 words