Science Magazine - January 6, 2017

A year ago, Science’s Editor-in-Chief Marcia McNutt highlighted two new journals in the Science fam- ily. Indeed, with the 2016 launches of Science Im- munology and Science Robotics, the Science fam- ily now has six members including, in addition, Science, Science Signaling, Science Translational Medicine, and Science Advances. This growth has occurred through a number of distinct opportunities, involving the emergence of new multidisciplinary research areas as well as alternative ways to accelerate the communication of research. The family now embod-ies a range of publishing and editorial models, with the content of Science Robotics and the open-access journal

Science Advances handled by academic editors. But the Science journals share a common focus on high quality and, in general, relatively broad scopes. This focus will be maintained as we consider adding to the family. We encourage enthusiasts and the curious to peruse our newest journal, Science Robotics, which is now freely available with a trial subscription through May 2017 to individuals and institutions with

Science site licenses.

How can we describe the contents of the Science family of journals and relate this profile to that of other journals? Of course, this can be most directly done by examining editorial statements and reading papers from the different journals. An alternative approach involves using natural language processing methods to analyze the contents as described in the accompanying post in Sciencehound (http://blogs.

sciencemag.org/sciencehound/2017/01/05/science-fam- ily-journal-content-analysis). Briefly, my approach involves taking abstracts from a large body of published papers, measuring the “distance” between each pair of papers based on the fraction of words that the abstracts have in common, weighting words that are rare in the overall body of papers, and then depicting these relative distances in a two-dimensional projection.

Assessing abstracts in this manner from 3 years of
Science, as well as most papers from the other Science

family journals published in 2016, results in a three-
pointed figure. Examination of the papers that lie in
each of the three arms reveals that one arm corresponds
primarily to biomedical sciences, one to physics and
physical sciences, and one to Earth sciences and evolu-
tion. That this relatively simple analysis generates an
intuitive structure relating the different disciplines is
perhaps not surprising. However, it is reassuring that
an unbiased analysis of the language that describes
scientific projects captures this essential feature of the
scientific enterprise.

What is the value of such an examination for the Sci-
ence family of journals? More detailed analysis reveals
that these journals cover different portions within each
arm, albeit with some overlap. Papers from Science and

Science Advances cover the landscape broadly with substantial representation in all three arms and in the central core. Those from

Science Signaling, Science Translational Medicine, and Science Immunology lie primarily in the biomedical arm. As expected, papers from the inaugural issue of

Science Robotics lie near the center extending toward the physical sciences arm. This raises interesting questions about the characteristics of papers that lie close to one another in the three-arm landscape and the relative outliers. Following trends over time using this kind of analysis could reveal aspects of the evolution of different disciplines and subdisciplines, and may even influence the launching of new journals or other modes of scientific communication. Applying analytical tools such as this natural language processing approach to journal content has considerable potential for assessing the structure of the scientific enterprise and the relationships of different disciplines. Gleaning knowledge from such data sets is an important goal; the challenge is to find the best questions and tools.

As the new year begins, I anticipate that many exciting manuscripts will be published in the Science family of journals. For active researchers, we hope that you will continue to consider our journals for your most exciting results. For all readers of our journals, I welcome feedback on all fronts.