Comparison of subject classification systems for resources in Physical Sciences Data Infrastructure (PSDI)

Comparison of subject classification systems for resources in Physical Sciences Data Infrastructure (PSDI) Dr Aileen Day 1st December 2025 1 Overview Physical Sciences Data Infrastructure (PSDI) is an integrated data infrastructure that directly supports researchers in the physical sciences in the UK to manage, transform and share their data. It is a project funded by the Engineering and Physical Sciences Research Council (EPSRC) and developed by the Sciences and Technology Facilities Council (STFC) and the University of Southampton. The first version of its outputs was released in spring 2025. The aim of this document is to review options to choose a dcat:themeTaxonomy (taxonomy/vocabulary/ontology/classification system) to describe the dcat:theme (subject/ research topic/ scientific discipline) of PSDI resource themes and resources in the PSDI resource catalogue. This will be used: 1. In the DCAT metadata which describes the catalogue of resources and resource themes https://metadata.psdi.ac.uk/psdi-dcat.jsonld 2. In the PSDI What We Provide resource catalogue to facilitate resource filtering and add tags to resource pages 3. In the PSDI Cross Data Search to facilitate data source filtering To avoid confusion with the PSDI term resource theme we will either refer to this field explicitly as dcat:theme or “subject” in this document (rather than just “theme”). While PSDI does capture dcat:theme values for its resources currently, this work is to investigate whether other classification system options offer any benefits over this, and to apply additional hierarchical grouping to guide users when filtering resources and datasets by subject. 1.1 Current subject classification in PSDI PSDI are currently using EuroSciVoc to classify subject of resources and resource themes: • in the jsonld representation of the PSDI resource catalogue: o dcat:theme in https://metadata.psdi.ac.uk/psdi-dcat.jsonld "dcat:theme": { "skos:prefLabel": "physical sciences", "@type": "skos:Concept", "@id": "http://data.europa.eu/8mn/euroscivoc/d3b09b78-ac5c-4fc3-b58a9573daf0e304" } • in the PSDI What We Provide website representation of the PSDI resource catalogue - in its filters which are populated by the jsonld representation. Currently, the subject filter simply displays the labels of these EuroSciVoc concepts themselves in a list without any of its hierarchy or grouping: Figure 1.1.2 Subject filter in PSDI resource themes webpage • in the resource webpages of the PSDI What We Provide resource catalogue the matching EuroSciVoc term is displayed with a link to its EuroSciVoc persistent identifier (PID): Figure 1.1.2 Example PSDI resource landing page showing linked dcat:theme terms (dcat:term values are linked to the PID for the term) Note that PSDI resource themes and resources are tagged according to the terms which describe them best, regardless of their level of the EuroSciVoc hierarchy. This work will enable the following improvements to PSDI: • ensuring that the classification system used is as intuitive as possible for resource owners to find relevant terms to tag their resources • grouping of subject filters in the PSDI What We Provide web pages (in Figure 1.1.2 above) – currently the exact terms that apply to the resources are shown in a list, without any hierarchy or grouping, which will become unwieldy as PSDI grows and there are more resources with more subjects • applying this grouping to enable filtering of data sets in the PSDI Cross Data Search. Figure 1.1.3 shows that only simple text filtering is applied currently. Grouped, hierarchical filtering of data sources (consistent with that in the PSDI resource catalogue) would be of benefit to the user. Figure 1.1.3 Current data source selection in the PSDI Cross Data Search 1.2 PSDI subject classification requirements The following are must-have requirements for a subject classification system for dcat:theme in PSDI: 1. Intuitive: the classification system needs to be straightforward for resource-owners and PSDI users to understand and navigate. 2. Physical Sciences scope: the classification system needs to include the whole scope of PSDI. We broadly align this with the Remit, programmes and priorities of the Engineering and Physical Sciences Research Council (EPSRC): chemistry; engineering; information and communications technologies; materials; mathematical sciences; physics, but in line with their policy some research areas that are slightly peripheral may also be included. 3. Maintained and versioned: the classification system will need to be updated as research and its focuses evolve and develop. This should be done in a managed way, so that the classification system as a whole needs to be versioned, as do individual terms - legacy terms should be preserved but flagged and any new terms should be linked to replaced terms where appropriate. 4. Hierarchical: some level of hierarchy and grouping into one or more levels make it easier to navigate to find applicable terms and can provide different levels of classification detail as required. The following are nice-to-have requirements for a classification system for dcat:theme in PSDI: 1. PIDs (persistent identifiers): terms that describe dcat:theme need to have a PID (to persistently and uniquely identify each term, ideally as an Internationalized Resource Identifier (IRI)) but this is indicated as “nice to have”, because if a chosen classification system does not have PIDs we could formalise an existing classification system into a PSDI SKOS Simple Knowledge Organization System Reference vocabulary with PIDs and maintain this ourselves. 2. Definitions: term definitions will make navigation and choice of terms clearer. Term definitions of subject terms can be useful in clarifying their scope but can be complex to write and interpret, so we have classed this as a nice-to-have requirement since if the subject term names or labels are chosen carefully, these can be easier to process. 3. Adoption: if the subject classification system is used in other organisations, repositories, reports, analyses etc. then it would help PSDI’s interoperability. This is classed as a nice-tohave requirement because mappings and cross-walks can be developed and used to increase interoperability (although we acknowledge that these can be complex and it is not always possible to define one-to-one relationships for all terms). 1.3 Subject classification selection process The process to select a subject classification to capture dcat:theme of PSDI resources (as described in this document) involved the following steps: • Review a wide range of subject classification candidates and for each provide a summary, an evaluation against the requirements above and outcomes or further actions, in particular whether it should be shortlisted (see section 2 All classification system candidates). • For candidates shortlisted in the previous step, PSDI version 1 resource themes and PSDI domain categories are mapped onto their terms as further evaluation of intuitiveness and suitability (see section 3 Mapping of PSDI resource themes onto shortlisted candidates). • Final ranking of shortlisted candidates and next steps (see section 4 Conclusions). 2 All classification system candidates We attempted to investigate a wide range of subject classification systems. For context, we start with the PSDI domain categories (an in-house categorisation of domains used by PSDI). Then, OECD (Organisation for Economic Co-operation and Development) FORD (Fields of Research and Development)/FOS (Fields of Study) is described since this is a widely accepted and adopted subject categorisation. This categorisation frames the top-level structure of EuroSciVoc (which is currently used to capture dcat:theme in PSDI) and here we describe why this was the first shortlisted candidate. The other two shortlisted candidates are described next: Modern Science Ontology (modsci), and OpenAlex Topics. We then move onto subject classifications which were not shortlisted but were reviewed. Like OECD FORD/FOS, the UNESCO (United Nations Educational, Scientific and Cultural Organization) nomenclature for fields of science and technology is commonly used worldwide to classify research and funding on a national and international level. MeSH (Medical Subject Headings) and NCI (National Cancer Institute) Thesaurus (NCIt) are vocabularies that are widely used for US life sciences public resource categorisation (and beyond). The next 5 subject classifications are used by funding bodies of various kinds to categorise research funded by them and its outputs: EPSRC (Engineering and Physical Sciences Council), ERC (European Research Council), DFG (Deutsche Forschungsgemeinschaft), Australian and New Zealand Standard Research Classification (ANZSRC), and the Royal Society. Nature are academic publishers who use classification systems to group their outputs on a journal and article level. The nfdiCoreOntology captures academic discipline on a high level and are a useful point of comparison, since it was developed by NFDI who are in many ways the German equivalent of PSDI (since they are the national infrastructure for scientific research data in Germany). Then as a final point of comparison, various higher education teaching and library classification systems were listed. 2.1 PSDI domain categories Figure 2.1.1 PSDI domain categories Summary: • Derived in-house by PSDI (see Appendix 1 for more details). • Categories of Physical Science derived from analysis of: Institution of Chemical Engineers (IChemE) special interest groups; Royal Society of Chemistry (RSC) interest groups; Institute of Physics (IOP) special interest groups; and Royal Society of Chemistry (RSC) journals. These were grouped together and assigned names as described in Appendix 1. • Used previously in PSDI to categorise funding applications and to ensure a wide representation of stakeholders for the PSDI advisory board. • See Appendix 2 for a mapping of PSDI resource themes (as of summer 2025) to these PSDI domain categories. Evaluation with respect to requirements: • Intuitive:  Simple and intuitive. The terms correspond to active communities in the physical sciences that researchers are familiar with. • Physical Sciences scope:  Yes – matches PSDI scope exactly. Quite chemistry focussed. • Maintained and versioned:  Not yet (PSDI would need to do this). • Hierarchical:  Single level with no hierarchy. To be more useful for PSDI, it would need a categorisation layer above it (to group the 18 categories into higher level disciplines) and more detailed field definitions within each of them. • PIDs (persistent identifiers):  Not yet. It would need to be converted into a SKOS vocabulary that is published and maintained by us and could be updated as required. • Definitions:  Not yet (PSDI would need to do this). This would help with some ambiguity in terms e.g. • “Chemistry Materials” makes sense when in context with “Engineering Materials” but the term isn’t commonly used otherwise. • The distinction between “Sustainability & Energy” and “Environmental” is subtle. • “Methods” is somewhat ambiguous and would need careful definition. • Adoption:  Not used by other repositories, so less interoperable than other candidates, unless mapping to them is implemented. Other considerations: •  This classification has been used by PSDI in the past and would provide a level of consistency with other PSDI processes and language. •  We would have full control over the terms and their definitions to fit to PSDI’s purposes. Outcome: • Not shortlisted because of the requirements that are not met. • If these PSDI domain categories are still regarded as useful in the future in parallel with the final chosen categorisation system, then the terms of both should be mapped. 2.2 OECD FORD (Fields of Research and Development)/Fields of Study (FOS) Non-exhaustive example snippet: Latest full list is available at: • “ANNEX 2 COMPARISON OF THE REVISED FOS CLASSIFICATION WITH THAT IN FM 2002”, page 12, of FOS.pdf which is available for download from Revised Field of Science and Technology (FOS) classification in the Frascati Manual - EconStatKB - UN Statistics Wiki). Summary: • The FORD (Fields of Research and Development) as defined by the Organisation for Economic Co-operation and Development (OECD) in the Frascati Manual, which was first published in 1963. • The latest (2015) version of the Frascati manual is OECD's 2015 Frascati Manual and in this: o The fields are given on page 59 in “Table 2.2. Fields of R&D classification” which lists 6 main “Broad Classification” values e.g. “1. Natural Sciences” with a column which lists “Second-level classification” values for each these e.g. “1.1 Mathematics”, “1.2 Computer and information sciences” etc. o This categorisation makes some interesting observations about fields of research or study categorisation: ▪ The complex classification process for research field is highlighted by the observation on page 58 of the manual that it is based on: • The knowledge sources drawn upon for the R&D activity carried out. • The objects of interest. • The methods, techniques and professional profiles of the scientists. • The areas of application. • 1. Natural Sciences • 1.1 Mathematics • 1.2 Computer and information sciences • 1.3 Physical sciences • 1.4 Chemical sciences • 1.5 Earth and related environmental sciences • 1.6 Biological sciences • 1.7 Other natural sciences • 2. Engineering and Technology • 2.1 Civil engineering • 2.2 Electrical engineering, electronic engineering, information engineering • 2.3 Mechanical engineering • 2.4 Chemical engineering • 2.5 Materials engineering • 2.6 Medical engineering • 2.7 Environmental engineering • 2.8 Environmental biotechnology • 2.9 Industrial Biotechnology • 2.10 Nano-technology • 2.11 Other engineering and technologies • 3. Medical and Health Sciences o … • 4. Agricultural Sciences o … • 5. Social Sciences o … • 6. Humanities o … o Another consideration when categorising research is outlined on page 45 which explains the difference between “Basic research”, “Applied research” and “Experimental development” in certain example fields (with examples in various fields in pages that follow). • The fields were updated in 2021 in Revised Field of Science and Technology (FOS) classification in the Frascati Manual - EconStatKB - UN Statistics Wiki. This document has some more granular classifications (but these are more illustrative examples rather than an exhaustive list with no identifiers for them). Evaluation with respect to requirements: • Intuitive:  simple and intuitive (the result of much research and thought): • The report describes the well-considered research and thought that went into the derivation of this classification system. • One limitation of this classification is that it uses the terms “physical sciences” and a separate neighbouring term “chemical sciences” which are broadly equivalent to “physics” and “chemistry” respectively (maybe with a slightly wider scope). “Physical sciences” defined in this way is a somewhat confusing category in the context of PSDI, where “physical sciences” is used as a broader term which includes “physics” and “chemistry”. Most resources in the Physical Sciences Data Infrastructure would not be categorised as “physical sciences”. • This classification system is not granular enough for use in PSDI directly (all version 1 PSDI resources would belong to only four categories: “1.3 Physical sciences”; “1.4 Chemical sciences”; “2.4 Chemical engineering”; “2.5 Materials engineering” or “2.10 Nano-technology”). • Physical Sciences scope:  Yes in breadth, but not granular enough. • Maintained and versioned:  Yes, updated periodically. Last released 2021. • Hierarchical:  Yes – 2 levels. • PIDs (persistent identifiers):  Not as such, but numbered for reference. • Definitions:  Not formal definitions (but some examples given in accompanying report). • Adoption:  Very well accepted and broadly adopted - used as the basis for many studies and reports e.g. from funders and governments. It is also used by Dryad https://datadryad.org/ to allow users to filter data to subjects of relevance, and DMPOnline, an online Data Management Plan creation tool developed by the Digital Curation Centre (DCC) so that researchers can pick what guidance, options and templates are available based on their specified research discipline. Outcome: • Not suitable for PSDI directly because it is not granular enough. 2.3 EuroSciVoc Non-exhaustive example snippet: Full list is available: • For download from http://data.europa.eu/8mn/euroscivoc/. • For browsing at https://op.europa.eu/en/web/eu-vocabularies/concept-scheme/- /resource?uri=http://data.europa.eu/8mn/euroscivoc/40c0f173-baa3-48a3-9fe6d6e8fb366a00. Summary: • Multilingual taxonomy of science fields based on OECD FORD (Fields of Research and Development) but extended in granularity through a semi-automatic process developed with Natural Language Processing (NLP) techniques with content from CORDIS (a repository of EU research results). Evaluation with respect to requirements: • Intuitive:  Yes on the whole but we note: • social sciences • … • natural sciences o earth and related environmental sciences ▪ … o mathematics ▪ … o biological sciences ▪ … o chemical sciences ▪ organic chemistry • … ▪ inorganic chemistry • … ▪ nuclear chemistry • … ▪ electrochemistry • … ▪ physical chemistry • thermochemistry • photochemistry • quantum chemistry ▪ polymer sciences • … ▪ catalysis • … ▪ analytical chemistry • … o physical sciences ▪ … • engineering and technology o … • humanities o … • agricultural sciences o … • medical and health sciences o … 2.6 UNESCO nomenclature for fields of science and technology Non-exhaustive example snippet: • 11 Logic o … • 12 Mathematics o … • 21 Astronomy and astrophysics o … • 22 Physics o … • 23 Chemistry o 2301 Analytical Chemistry ▪ 2301.01 Absorption spectroscopy ▪ 2301.02 Biochemical analysis ▪ 2301.03 Chromatographic analysis ▪ 2301.04 Electrochemical Analysis ▪ 2301.05 Emission spectroscopy ▪ 2301.06 Fluorimetry ▪ 2301.07 Gravimetry ▪ 2301.08 Infrared spectroscopy ▪ 2301.09 Magnetic resonance spectroscopy ▪ 2301.10 Mass spectroscopy ▪ 2301.11 Microchemical analysis ▪ 2301.12 Microscopy ▪ 2301.13 Microwave spectroscopy ▪ 2301.14 Phosphorimetry ▪ 2301.15 Polymer analysis ▪ 2301.16 Radiochemical analysis ▪ 2301.17 Raman spectroscopy ▪ 2301.18 Thermal analytical methods ▪ 2301.19 Volumetry ▪ 2301.20 X-Ray spectroscopy ▪ 2301.99 Other (specify) o 2302 Biochemistry ▪ … o 2303 Inorganic chemistry ▪ … o 2304 Macromolecular chemistry ▪ … o 2305 Nuclear Chemistry ▪ … o 2306 Organic Chemistry ▪ … o 2307 Physical chemistry ▪ … o 2390 Pharmaceutical chemistry ▪ … o 2399 Other chemical specialties (specify) ▪ … • 24 Life Sciences o … • 25 Earth and Space Sciences o … • 31 Agricultural Sciences o … • 32 Medical Sciences o … • 33 Technological Sciences o … • 51 Anthropology o … • … Full list is available at: • SKOS: Download (2007 version) for download in Turtle or RDF format. • or can be browsed at Hierarchical Browsing of SKOS: UNESCO nomenclature for fields of science and technology. Summary: • Developed by UNESCO (United Nations Educational, Scientific and Cultural Organization). • According to SKOS: UNESCO nomenclature for fields of science and technology: o The Proposed international standard nomenclature for fields of science and technology was proposed in 1973 and 1974 by the Division of Science Policy and Statistics for Science and Technology of UNESCO and adopted by the Scientific Advisory Committee. This is a classification system widely used in knowledge management of research projects and dissertations. Categories are divided into three hierarchical levels: ▪ Fields: Referring to general sections. Encoded with two digits and comprises several disciplines. ▪ Disciplines: Provide an overview of specialty groups in Science and Technology. Encoded with four digits. Despite being different from each other disciplines with cross references, or within the same field, are considered to have common characteristics. ▪ Subdisciplines: Entries are the more specific elements of the nomenclature and represent the activities that take place within a discipline. Encoded with six digits. In turn, must correspond to individual specialties in science and technology. o There is top-level 2-digit grouping, then 4-digit grouping within this then 6-digit within that. PSDI resources would need to be mapped down to the 6-digit level of granularity. o Multi-hierarchical - Physical chemistry is in it twice – under physics and under chemistry. Evaluation with respect to requirements: • Intuitive:  Yes with limitations: • Top level seems more intuitive than EuroSciVoc. • Very physics-based for subjects which PSDI might think of as chemistry (very limited range of chemistry subjects). • Physical Sciences scope:  Yes. • Maintained and versioned:  No – does not look like it’s been updated since 1970’s. • Hierarchical:  Yes – 2 digit, 4 digit and 6 digit levels of classification. • PIDs (persistent identifiers):  Yes. • Definitions:  No. • Adoption:  Widely used for classification of research papers and doctoral dissertations. Outcome: • Not shortlisted because it is not being updated, and its limited description of chemistry compared to EuroSciVoc (although we could learn from its arrangement of top-level subjects which are easier to navigate). 2.7 MeSH (Medical Subject Headings) Non-exhaustive example snippet: Full list is available at: • https://www.nlm.nih.gov/mesh/meshhome.html for download. • https://meshb.nlm.nih.gov/treeView for browsing. Summary: • https://www.nlm.nih.gov/mesh/meshhome.html says “The Medical Subject Headings (MeSH) thesaurus is a controlled and hierarchically-organized vocabulary produced by the National Library of Medicine. It is used for indexing, cataloging, and searching of biomedical and healthrelated information. MeSH includes the subject headings appearing in MEDLINE/PubMed, the NLM Catalog, and other NLM databases.”. Evaluation with respect to requirements: • Intuitive:  The disciplines that are included are organised intuitively, but there are omissions due to the life-science focus of mesh (see notes about Scope). • Physical Sciences scope:  Does not cover the whole of the Physical Sciences scope of PSDI: • It is not very granular (e.g. “Materials Science [H01.413]” and “Chemistry, Organic [H01.181.404]” are not broken down into more detail. • “Engineering” is not included. • Maintained and versioned:  Yes. • … • Disciplines and Occupations [H] o Natural Science Disciplines [H01] ▪ Biological Science Disciplines [H01.158] ▪ Chemistry [H01.181] • Biochemistry [H01.181.122] o … • Cheminformatics [H01.181.169] • Chemistry, Agricultural [H01.181.216] • Chemistry, Analytic [H01.181.309] • Chemistry, Clinical [H01.181.341] • Chemistry, Inorganic [H01.181.370] o … • Chemistry, Organic [H01.181.404] • Chemistry, Pharmaceutical [H01.181.466] • Chemistry, Physical [H01.181.529] o Crystallography [H01.181.529.240] o Electrochemistry [H01.181.529.307] o Photochemistry [H01.181.529.711] o Radiochemistry [H01.181.529.776] • Computational Chemistry [H01.181.590] • Microchemistry [H01.181.650] ▪ Earth Sciences [H01.277] ▪ Environmental Science [H01.345] ▪ Materials Science [H01.413] ▪ Mathematics [H01.548] ▪ Microtechnology [H01.570] ▪ Nanotechnology [H01.603] ▪ Physics [H01.671] ▪ Science [H01.770] o Health Occupations [H02] • … • Hierarchical:  Yes. • PIDs:  Yes. • Definitions:  Yes. • Adoption:  Yes used widely in the life sciences, and North America. Outcomes: • Not shortlisted because the wider context of MESH tends to skew its focus to life sciences rather than physical sciences, it does not cover the full scope of PSDI and does not describe it in sufficient detail for our purposes. 2.8 NCI (National Cancer Institute) Thesaurus (NCIt) Non-exhaustive example snippet: Full list is available at: • https://evsexplore.semantics.cancer.gov/evsexplore/hierarchy/ncit/C19160 for browsing “Occupation or Discipline ( Code - C19160 )”. Summary: • Thesaurus from NCI (National Cancer Institute). • C16328 Behavioral Sciences o … • C16344 Biological Sciences o … • C15178 Complementary and Alternative Medicine o … • C84341 Discovery Science o … • C26037 Food Safety o … • C19199 Health Sciences o … • C157884 Integrative Medicine o … • C45427 Medical Science o … • C148247 Natural and Applied Sciences o … • C25193 Occupation o … • C16987 Physical Sciences o C16414 Chemistry ▪ C16415 Analytical Chemistry ▪ C18470 Computational Chemistry ▪ C18794 Fiber Chemistry ▪ C16713 Immunochemistry ▪ C16418 Inorganic Chemistry ▪ C16419 Organic Chemistry ▪ C16421 Physical Chemistry • C19130 Chemical Dynamics • C18122 Chemical Kinetics • C16985 Photochemistry • C19048 Structural Chemistry ▪ C18188 Stereochemistry ▪ C64351 Surface Chemistry ▪ C20609 Synthesis Chemistry o C16633 Geography ▪ … o C16825 Mathematics ▪ … o C16989 Physics ▪ … • C19491 Population Sciences o … • C17141 Social Sciences o … • C17187 Technology o … • As described at that webpage: “NCI Thesaurus (NCIt) provides reference terminology for many NCI and other systems. It covers vocabulary for clinical care, translational and basic research, and public information and administrative activities.”. • Scientific subjects (stored under Occupation or Discipline ( Code - C19160 ) are only a small part of this thesaurus – it contains a wide range of other concepts and definitions used in life sciences. • Most PSDI resources fall under the category “Physical Sciences” but some will also fall under its adjacent category: “Technology” (under its subcategories “Engineering”, “Nanotechnology”, “Computer Sciences”, “Food Science and Technology”). “Engineering” for example contains subjects such as: “Bioengineering”, “Chemical Engineering”, “Electrical Engineering”, “Environmental Engineering”. • More granular than MeSH and more of an overlap with PSDI’s scope Evaluation with respect to requirements: • Intuitive:  Yes. • Physical Sciences scope:  Better representation of the Physical Sciences than MESH but focus is biomedical, and some key PSDI areas are not well represented e.g. C16418 - Inorganic Chemistry only contains one sub-category C16416 - Bioinorganic Chemistry. • Maintained and versioned:  Yes. • Hierarchical:  Yes. • PIDs:  Yes. • Definitions:  Yes. • Adoption:  Yes used widely in the life sciences, and North America. Outcomes: • Not shortlisted because its wider context skews its focus to life sciences rather than physical sciences, so it is lacking in detail about some areas of interest to PSDI. 2.9 EPSRC (Engineering and Physical Sciences Council) research areas Non-exhaustive example snippet: Full list is available at: • https://www.ukri.org/wp-content/uploads/2023/02/EPSRC-140223-ResearchAreasList1.pdf. • And https://www.ukri.org/what-we-do/browse-our-areas-of-investment-andsupport/?category=epsrc. Summary: • Developed by EPSRC (Engineering and Physical Sciences Research Council) to categorise research funded by them. • EPSRC research strategies are informed by a range of evidence sources: https://www.ukri.org/wp-content/uploads/2022/06/EPSRC-13062022-EPSRC-EvidenceSources-amended.docx. Evaluation with respect to requirements: • Intuitive:  Very difficult to find relevant terms because there are 108 terms with no organisation of hierarchy, listed in alphabetical order. • Physical Sciences scope:  Yes – matches PSDI scope exactly. • Maintained and versioned:  This categorisation is primarily focussed on funding policy so there is a risk that when it is updated it may be completely different. There is no guarantee that it will be of a similar form with any analogous, versioned, mapped terms. • Hierarchical:  Single level with no hierarchy. • PIDs:  No PIDs – not semantic and no identifier numbers. • Definitions:  Yes. • Algebra • Analytical science • Antihydrogen • Architectures and operating systems • Artificial intelligence technologies • Assistive technology, rehabilitation and musculoskeletal biomechanics • Bioenergy • Biological informatics • Biomaterials and tissue engineering • Biophysics and soft matter physics • Built environment • Carbon capture and storage • Catalysis • Chemical biology and biological chemistry • Chemical reaction dynamics and mechanisms • Clinical technologies (excluding imaging) • Coastal and waterway engineering • Cold atoms and molecules • Combustion engineering • Complex fluids and rheology • Computational and theoretical chemistry • Condensed matter: electronic structure • Condensed matter: magnetism and magnetic materials • Continuum mechanics • Control engineering • … • Adoption:  Yes – used in a different context for EPSRC research categorisation – reporting and grant allocation. Researchers who are PSDI contributors and users may have used this classification already when applying for grants. Other considerations: •  Matches scope and funding of PSDI - synergy with PSDI funding. Outcomes: • Not shortlisted because of the work that would be required to convert it into a semantic form with PIDs, especially considering that not all requirements have been met and the high risk that this effort would need to be repeated in the future if the EPSRC categorise their research in a different way. 2.10 ERC (European Research Council) panel structure for Physical Sciences and Engineering Non-exhaustive example snippet: Full list is available at: • defined in: https://erc.europa.eu/sites/default/files/document/file/ERC_Panel_structure_2021_2022.pdf. Summary: • Developed by European Research Council (ERC) to categorise research funded by them. Evaluation with respect to requirements: • Intuitive:  Yes - very comprehensive and detailed. • Physical Sciences scope:  Yes. • Physical Sciences and Engineering o PE1 Mathematics ▪ … o PE2 Fundamental Constituents of Matter ▪ … o PE3 Condensed Matter Physics ▪ … o PE4 Physical and Analytical Chemical Sciences ▪ PE4_1 Physical chemistry ▪ PE4_2 Spectroscopic and spectrometric techniques ▪ PE4_3 Molecular architecture and Structure ▪ PE4_4 Surface science and nanostructures ▪ PE4_5 Analytical chemistry ▪ PE4_6 Chemical physics ▪ PE4_7 Chemical instrumentation ▪ PE4_8 Electrochemistry, electrodialysis, microfluidics, sensors ▪ PE4_9 Method development in chemistry ▪ PE4_10 Heterogeneous catalysis ▪ PE4_11 Physical chemistry of biological systems ▪ PE4_12 Chemical reactions: mechanisms, dynamics, kinetics and catalytic reactions ▪ PE4_13 Theoretical and computational chemistry ▪ PE4_14 Radiation and Nuclear chemistry ▪ PE4_15 Photochemistry ▪ PE4_16 Corrosion ▪ PE4_17 Characterisation methods of materials ▪ PE4_18 Environment chemistry o PE5 Synthetic Chemistry and Materials ▪ … o PE6 Computer Science and Informatics ▪ … o PE7 Systems and Communication Engineering ▪ … o PE8 Products and Processes Engineering ▪ … o PE9 Universe Sciences ▪ … o PE10 Earth System Science ▪ … o PE11 Materials Engineering ▪ … • Life Sciences o … • Social Sciences and Humanities o … • Maintained and versioned:  This categorisation is primarily focussed on funding policy so when it is updated it may be significantly different. There is no guarantee that it will be of a similar form with any analogous, versioned, mapped terms. • Hierarchical:  Yes – 2 levels. • PIDs:  Numbered identifiers but not semantic – PSDI would need to convert this into a SKOS vocabulary. • Definitions:  No. • Adoption:  Yes – used in a different context for EU research categorisation – reporting and grant allocation. Researchers who are PSDI contributors and users may have used this classification already when applying for grants (although maybe less common in the UK than EPSRC funding). Outcomes: • Not shortlisted because of the work needed to maintain it in semantic form and because it doesn’t show sufficient improvement on other options to do this. 2.15 “academic discipline” in nfdiCoreOntology Non-exhaustive example snippet: Full list is available at: • “academic discipline” class within https://ise-fizkarlsruhe.github.io/nfdicore/. Summary: • Part of NFDIcore Ontology which is a mid-level BFO ontology for representing metadata related to NFDI (Nationale Forschungsdateninfrstrukur), including individuals, organizations, projects, data portals, and more. • It is somewhat higher level than we are looking for in PSDI - it is designed to link to more domain-specific ontologies which extend its core structure in a modular fashion e.g. The MatWerk ontology for Materials Science imports “academic discipline” and adds a class below it “research topic” but these are not enumerated. Evaluation with respect to requirements: • Intuitive:  Yes. • Physical Sciences scope:  Yes but at a high level – breakdown is not granular enough. • Maintained and versioned:  Yes. • Hierarchical:  It only has one level (which is very high level), but it is of modular design and intended to be extendable below this via linked domain-level ontologies to include more detail. • PIDs:  Yes. • Definitions:  Yes. • Adoption:  Yes at NFDI. Outcome: • Not shortlisted because there are alternatives which capture more granularity which is directly integrated within them. • agriculture • artificial intelligence • biology • chemistry • computer science • construction engineering • cybersecurity • data science • electrical engineering • geoscience • humanity • materials science • mathematics • mechanical engineering • medicine • natural science • physics • process engineering • social science • software engineering • theoretical computer science 2.16 Higher education teaching classification systems The following teaching subject classifications were not considered in detail because higher education teaching is not the same as research (and could show some lag in new areas of research), but their grouping systems were considered: • HECOS (Higher Education Classification of Subjects) and Common Aggregation Hierarchy (CAH) from HESA (Higher Education Statistics Agency) • JACS (Joint Academic Coding System) Principal subject codes from HESA (Higher Education Statistics Agency) (replaced by HECOS/CAH). • Classification of Instructional Programs (U.S. Department of Education's National Center for Education Statistics (NCES)). • Australian Standard Classification of Education (Australian bureau of statistics) • HESA (Higher Education Statistics Agency) classify academic staff according to their departments as described by Cost centres (2012/13 onwards) 2.17 Library classification systems The following library classifications were not considered in detail because library content is not the same as research (and could show some lag in new areas of research), but their grouping systems were considered: • Library of Congress Classification • Dewey Decimal Classification • Cambridge University Library Classification 3 Mapping of PSDI resource themes onto shortlisted candidates This extensive review of how subjects are described and classified in section 2 was useful not only in the selection of candidates, but also to observe recurring successful themes in their organisation. In particular, we observed that the general organisation system below was common and very intuitive to follow: - domain (e.g. Physical Sciences, Natural Sciences, Applied Sciences etc.) o subject discipline (e.g. Chemistry, Physics) ▪ sub-discipline (e.g. Organic Chemistry, Inorganic Chemistry) • field (more specific description of field). In PSDI it is not necessary to include the top-level since our scope generally corresponds to one subject grouping – the Physical Sciences. Hence, in our mapping process described in this section we also illustrate how for each candidate, a subset of the hierarchical grouping can be simplified to a consistent, 2-level hierarchy with subject discipline at the top-level and sub-disciplines within this (a recurring and sensible structure) which then contain the more specific fields. This section shows the mapping of PSDI resource themes (or resources) onto each shortlisted categorisation system. This allows us to evaluate requirement 1 (intuitive) in more detail. Note that: • We only include resource themes (and resources) which were in PSDI as of August 2025. • Mapping is performed at the resource theme level and can be assumed to apply to all resources within them, unless a resource has a distinct theme to its parent resource theme in some way (usually more specific) in which case they are included explicitly. • For hierarchical classification systems, a resource theme is assigned to the most relevant term, regardless of the level of the hierarchy. For some resource themes this will be a general term near the top of the hierarchy e.g. “chemistry” but for some it will be a more specific term e.g. at the most granular level of the classification. • We have simplified the hierarchies shown to a maximum of 2-levels to reflect the filters that will be shown in the PSDI “What we provide” pages (more levels become unwieldy and complicated for users). If the exact term that is relevant for a PSDI resource theme is more granular than level 2, the dcat:theme column indicates the term that would be shown on the resource/resource theme page itself (only Level 1 and 2 would show in the PSDI “What We Provide” filters though, and not this dcat:theme). • This simplified hierarchy follows the general structure: o Level 1 corresponds to discipline. o Level 2 corresponds to sub-discipline. o dcat:theme corresponds to specific term (if not one of the above). • It is possible for a resource or resource theme to appear more than once, if multiple subject terms are appropriate. • If a resource theme is not specific to a particular dcat:theme value, or defined set of dcat:themes, they are indicated as “Multidisciplinary”. • We have also simplified the hierarchies shown to only include the subset of classifications which are relevant to PSDI. Where required, we will add the top Level 1 term “Other” to contain terms which do not correspond to those explicitly shown for Level 1 (to simplify the options presented to the user when searching). Again, this is just for the purposes of the folder structure of the subject filters, and the dcat:theme values of these resource themes would still the properly defined term. There should not be many of these, since they would be at the edges of PSDI’s scope. The mapping shown in the subsequent sections are outlined in the accompanying file PSDI_Subject_Shortlist_Mappings.xlsx which lists: • “PSDI resource themes” – the PSDI resource themes and the URLs that describe them as of August 2025 which were mapped in this exercise • “PSDI_Domains_derivation” – the list of PSDI domains described in section 2.1 and details of their derivation • “1_EuroSciVoc” – mapping described in section 3.1. • “2_ModSci” – mapping described in section 3.2. • “3_OpenAlex” – mapping described in section 3.3. 3.1 Mapping of PSDI resource themes and PSDI Domain categories onto EuroSciVoc categories Level 1 Level 2 dcat:theme PSDI Domains PSDI resource themes earth and related environmental sciences environmental sciences - Environmental mathematics biological sciences molecular biology - Chemical Biology - BioSim (Biomolecular Simulations) Data Resources biochemistry - Biochemistry chemical sciences - PSDI Data Revival - Data Collections for Chemical Modelling - PSDI Data Conversion organic chemistry - Chemistry Materials - Organic & Pharmaceutical - Data Sources for PSDI Cross Data Search - Chemical Availability Search (ChASe); Propersea (Property Prediction); Chemotion Repository - Cambridge Structural Database (CSD) inorganic chemistry - Chemistry Materials - Thematic Portal (Catalysis Data Infrastructure Example) - Galaxy - Example Galaxy XAFS RO-Crate Workflows organometallic chemistry - Metal Organic - Cambridge Structural Database (CSD) physical chemistry - Physical Chemistry - Data Collections for Chemical Modelling - Physical Chemistry Properties Data Collection quantum chemistry - Collaborative Computational Project for NMR Crystallography (CCP-NC) polymer sciences - Polymers - Chemistry Materials Catalysis - Catalysis - Chemistry Materials - Thematic Portal (Catalysis Data Infrastructure Example) - Galaxy - Example Galaxy XAFS RO-Crate Workflows analytical chemistry - Analytical Chemistry physics (physical sciences) condensed matter physics - Solid-state Physics - Collaborative Computational Project for NMR Crystallography (CCP-NC) - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation soft matter physics - BioSim (Biomolecular Simulations) Data Resources Optics microscopy nonlinear optics - Case Studies: Discovering Value in Legacy Data - Resurrecting Second Harmonic Generation (SHG) Case Study absorption spectroscopy - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation computer and information sciences - Computational artificial intelligence machine learning - Data to Knowledge Software software development - PSDI Skills4Scientists - Python I; Python II; Version Control and Github simulation software - BioSim (Biomolecular Simulations) Data Resources - Collaborative Computational Project for NMR Crystallography (CCP-NC) data science data exchange - PSDI Data Conversion - PSDI Digital Lab Notebook Resources - PSDI Data Revival data processing - PSDI Digital Lab Notebook Resources - Case Studies: Discovering Value in Legacy Data - Galaxy computational science - Data Collections for Chemical Modelling engineering and technology chemical engineering - Chemical Engineering other engineering and technologies - Food environmental engineering - Sustainability & Energy nanotechnology - Nanotech materials engineering - Engineering Materials crystals - OPTIMADE Data Providers - Cambridge Structural Database (CSD) other multidisciplinary - Methods - Case Studies: Discovering Value in Legacy Data - Container Image Registry - Data Sources for PSDI Cross Data Search - Galaxy - PSDI Digital Lab Notebook Resources - PSDI Knowledge Base Topics - PSDI Skills4Scientists - Thematic Portal (Catalysis Data Infrastructure Example) Note that: • The level 1 classifications broadly correspond to the NT1 level of EuroSciVoc, and level 2 to NT2, apart from “engineering and technology” which is a top-level OECD FORD/FOS term (and its sub-disciplines are taken at the NT1 level). • We have renamed “physical sciences” to “physics” to avoid confusion in PSDI’s context. • We have adjusted some of these EuroSciVoc term assignments for resources compared to that published in PSDI currently in light of this review. • Edge cases to highlight: o “OPTIMADE Data Providers” and “Cambridge Structural Database (CSD)” moved from “crystallography” (under “geology”) to “crystals” (under “materials engineering”). o There is no “computer model” or “simulation” term for e.g. “Data Collections for Chemical Modelling” and many other resources. 3.2 Mapping of PSDI resource themes and PSDI Domain categories onto ModSci categories Level 1 Level 2 dcat:theme PSDI Domains PSDI resource themes Computer Science - Computational Artificial Intelligence Machine Learning - Data to Knowledge modsci#ComputerApplications - Container Image Registry modsci#ComputerSoftware - PSDI Skills4Scientists - Python I; Python II; Version Control and Github - BioSim (Biomolecular Simulations) Data Resources modsci#DataFormat - PSDI Data Conversion - PSDI Data Revival - Case Studies: Discovering Value in Legacy Data - PSDI Digital Lab Notebook Resources Engineering Mathematics Scientific Modelling - Computational - BioSim (Biomolecular Simulations) Data Resources - Collaborative Computational Project for NMR Crystallography (CCP-NC) - Data Collections for Chemical Modelling - Data to Knowledge - Galaxy Biology Biochemistry - Biochemistry Molecular Biology - Chemical Biology - BioSim (Biomolecular Simulations) Data Resources Chemistry - PSDI Data Revival - Data Collections for Chemical Modelling - PSDI Data Conversion Analytical Chemistry - Analytical Chemistry Biochemistry - Biochemistry Chemical Engineering - Chemical Engineering Crystallography - OPTIMADE Data Providers - Cambridge Structural Database (CSD) Environmental Chemistry - Environmental Food Chemistry - Food Green Chemistry - Environmental - Sustainability & Energy Inorganic Chemistry - Thematic Portal (Catalysis Data Infrastructure Example) - Galaxy - Example Galaxy XAFS RO-Crate Workflows Materials Science - Chemistry Materials - Engineering Materials modsci_Nanochemistry - Nanotech modsci_Polymerisation Mechanisms - Polymers modsci_TheoryAndDesig nOfMaterials - OPTIMADE Data Providers Organic Chemistry - Organic & Pharmaceutical - Cambridge Structural Database (CSD) - Data Sources for PSDI Cross Data Search - Chemical Availability Search (ChASe); Propersea (Property Prediction); Chemotion Repository Organometallic Chemistry - Metal Organic - Cambridge Structural Database (CSD) - Galaxy - Example Galaxy XAFS RO-Crate Workflows - Thematic Portal (Catalysis Data Infrastructure Example) Physical Chemistry - Physical Chemistry - Data Collections for Chemical Modelling - Physical Chemistry Properties Data Collection - Collaborative Computational Project for NMR Crystallography (CCP-NC) modsci_Catalysis - Catalysis - Galaxy - Example Galaxy XAFS RO-Crate Workflows Earth Science Environmental Science - Environmental Physics Condensed Matter Physics - Solid-state Physics - Collaborative Computational Project for NMR Crystallography (CCP-NC) - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation modsci_OpticalPhysics modsci_NonlinearOptics - Case Studies: Discovering Value in Legacy Data - Resurrecting Second Harmonic Generation (SHG) Case Study modsci_Spectroscopy - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation other multidisciplinary - Methods - Case Studies: Discovering Value in Legacy Data - Container Image Registry - Data Sources for PSDI Cross Data Search - Galaxy - PSDI Digital Lab Notebook Resources - PSDI Knowledge Base Topics - PSDI Skills4Scientists - Thematic Portal (Catalysis Data Infrastructure Example) Note that: • The Level 1 classifications broadly correspond to the 2nd level under Science and Level 2 to the level directly under that. • Edge cases to highlight: o The definition for Physical Chemistry mentions “quantum chemistry” but this is not listed under it so e.g. “Collaborative Computational Project for NMR Crystallography (CCP-NC)” needs to be tagged with this more general term. 3.3 Mapping of PSDI resource themes and PSDI Domain categories onto OpenAlex categories Level 1 Level 2 dcat:theme PSDI Domains PSDI resource themes Agricultural and Biological Sciences Food Science - Food Biochemistry, Genetics and Molecular Biology Biochemistry - Biochemistry Molecular Biology - Chemical Biology Protein Structure and Dynamics - BioSim (Biomolecular Simulations) Data Resources Chemical Engineering - Chemical Engineering Bioengineering Nanotechnology research and applications - Nanotechnology Catalysis - Catalysis - Galaxy - Example Galaxy XAFS RO-Crate Workflows - Thematic Portal (Catalysis Data Infrastructure Example) - Catalysis Data Infrastructure (CDI) Database Chemistry - Data Collections for Chemical Modelling - PSDI Data Conversion - PSDI Data Revival Analytical Chemistry - Analytical Chemistry Electrochemistry Inorganic Chemistry Crystal structures of chemical compounds - OPTIMADE Data Providers - Cambridge Structural Database (CSD) Zeolite Catalysis and Synthesis - Galaxy - Example Galaxy XAFS RO-Crate Workflows - Thematic Portal (Catalysis Data Infrastructure Example) - Catalysis Data Infrastructure (CDI) Database Organic Chemistry - Organic & Pharmaceutical - Cambridge Structural Database (CSD) - Data Sources for PSDI Cross Data Search - Chemical Availability Search (ChASe), Propersea (Property Prediction) Organometallic Complex Synthesis and Catalysis - Galaxy - Example Galaxy XAFS RO-Crate Workflows - Thematic Portal (Catalysis Data Infrastructure Example) - Catalysis Data Infrastructure (CDI) Database Organometallic Compounds Synthesis and Characterization - Metal Organic - Cambridge Structural Database (CSD) Chemical Thermodynamics and Molecular Structure - Data Collections for Chemical Modelling - Physical Chemistry Properties Data Collection Physical and Theoretical Chemistry - Physical Chemistry Advanced Physical and Chemical Molecular Interactions thermodynamics and calorimetric analyses - Data Collections for Chemical Modelling - Physical Chemistry Properties Data Collection Spectroscopy - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation Advanced NMR Techniques and Applications - Collaborative Computational Project for NMR Crystallography (CCP-NC) Computer Science - Computational Artificial Intelligence Machine Learning and Algorithms - Data to Knowledge Computational Theory and Mathematics Modeling and Simulation Systems - Data Collections for Chemical Modelling - Collaborative Computational Project for NMR Crystallography (CCP-NC) - BioSim (Biomolecular Simulations) Data Resources Computer Science Applications Teaching and Learning Programming - PSDI Skills4Scientists - Python I; Python II; Version Control and Github Information Systems Advanced Technology in Applications Research Data Management Practices - PSDI Data Conversion - PSDI Data Revival - PSDI Digital Lab Notebook Resources - PSDI Skills4Scientists - Case Studies: Discovering Value in Legacy Data - Galaxy Earth and Planetary Sciences Energy - Sustainability & Energy Engineering Mechanics of Materials - Engineering Materials Muon and positron interactions and applications - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation Environmental Science - Environmental Materials Science Electronic, Optical and Magnetic Materials Nonlinear Optical Materials Research - Case Studies: Discovering Value in Legacy Data - Resurrecting Second Harmonic Generation (SHG) Case Study Materials Chemistry - Chemistry Materials - OPTIMADE Data Providers Mesoporous Materials and Catalysis - Galaxy - Example Galaxy XAFS RO-Crate Workflows - Thematic Portal (Catalysis Data Infrastructure Example) - Catalysis Data Infrastructure (CDI) Database Machine Learning in Materials Science - Data to Knowledge Polymers and Plastics - Polymers Mathematics Physics and Astronomy Condensed Matter Physics - Solid-state Physics - Collaborative Computational Project for NMR Crystallography (CCP-NC) - Galaxy Training Network: Finding the Muon Stopping Site; MuDirac Documentation; MuSpinSim Documentation; Xray Larch Documentation Theoretical and Computational Physics