分类:Scientometrics

Scientometrics studies development of science, activities of scientists, and diffusion and communication of science, and aims to help development of science, to become and educate better scientists and the general public, to form a better basis of administration of science development. For that, we need to have a good understanding of the objects of study and have good mathematical models/patterns of the behaviors of the objects of study.

Then what are the typical data, typical questions asked in scientometrical studies, typical ways of thinking, typical methods of analysis in scientomtrics, and how this discipline is related to other disciplines and serves the whole society? For more discussion on these issue, please read [什么是科学计量学] on Jinshan's Blog site [吴金闪的工作和思考]. For now it is in Chinese and hopefully soon it will be rewritten in English.

Besides as a discipline of its own, scientometrics can also be regarded as an example of Data Science: scientometrics provides good data source, questions of study, driven force of methodological development and test bed of methods for data science. All these studies in the future might be applicable to other systems of data science.

We have an ambitious goal about research work in scientometrics. We are looking forward to making a difference to the field from the following several aspects. One should be able to see from the following that we are trying to make this field more like natural science, more coherent especially mathematically, and closer to real-world problems.

Firstly, we propose to use multi-layer network as a general framework to represent data and typical questions, and also the basis to develop methods of analysis which naturally can incorporate typical ways of thinking and typical methods of analysis in network science. More specifically, for scientific papers, we can use an author-paper-concept network and for patents, we can use an inventor-patent-technology network, or the combined publication-patent network for both. Furthermore, when possible, we can add one more layer on products (development, producing and consumption of products, which is called in Economics as Input-Output Analysis into the network so that it becomes a publication-patent-economics network.

Secondly, once presented in a network from, the network effect, which means to consider both direct and indirect connections among nodes in networks, should naturally play an essential role in scientometrics. Examples of such consideration can be found in for instance [PageRank] and General Input-Output Analysis.

Thirdly, via the layer of concepts and the connection between papers and concepts, content-based data can be incorporated into scientometrical studies. For example, by locating each paper on the concept map, it should be much easier to see the main contribution of the paper (on this issue, by the way, we should work out the concept map of scientometrics as an example and use this to generate a sub concept map for each paper and publish this sub map together with the paper as an additional abstract for those several journals on scientometrics). Moreover, it might also be able to help do a much better job on classifications of papers, or on measuring creativity of scientific research work。

Lastly, via this new framework, we should be able to get more reliable data for scientometrical studies, for example at identifying authors and identifying citation backbone.

In all and all, we believe that multi-layer network is the infrastructure of scientometrics, a coherent way to present and refine the data and questions, a framework to develop and test methods of analysis.

One day, after some good progresses have been made in this direction, we are planing to write a textbook on scientometrics: Invitation to Scientometrics and of course also review papers, focusing on the big picture of scientometrics: typical objects of studies, typical questions, typical ways of thinking, typical methods of analysis and its relation to other disciplines and the real world, and of course, emphasizing the general framework.