In most of our bibliometric studies, we recommend using our classification of science, developed in-house by our experienced analysts. This classification of research outputs categorizes scientific journals and articles by S&T domain, field and subfield (5 domains, 20 fields and 174 subfields), with the subfields being mutually exclusive. We developed the first iteration of this classification scheme in 2010 in the course of a contract for the European Commission. It is based on best-practice taxonomies and was created in response to the lack of a widely adopted single classification scheme within the international bibliometric community. We modelled the categories on those of existing journal classifications (e.g., ISI, CHI, ERA), and used their groupings of journals as “seeds” or attractors for journals in the new classification. We assigned individual journals to single, mutually exclusive categories via an approach combining algorithmic methods and expert judgment. We also designed the classification to be as inclusive as possible of newer fields of inquiry, general and multidisciplinary journals, and the range of arts and humanities disciplines. This classification scheme is now well established, and we have used it for projects investigating a variety of scientific domains and geographic regions. Furthermore, it can easily be mapped to many other classifications of science, such as the Medical Subject Headings developed by the United States National Library of Medicine.

In 2019, in the course of a contract for SRI International, we further developed the journal-based classification into the article-level and hybrid versions. For the article-level classification, a scientific publication is attributed to a domain, field and subfield based on its title, abstract, keywords, author affiliation and the classification of its citations, using a deep neural network (an artificial intelligence technique). In the hybrid version, most articles are still classified at the journal level, except for those published in multidisciplinary journals (e.g., Science, Nature, PNAS and PLOS One), which are classified at the article level.

Read more about the development of the article-level classification here and the journal-level classification here

The journal-level classification is freely available under a creative commons license and is operational in 26 languages. Science-Metrix encourages the use of this tool in any research, education and librarianship endeavors. Furthermore, feedback is essential to the continued improvement of the classification and is always welcome.

Download the classification [Excel file].