How can I access research classification systems and/or research topic analyses?
Answer
Understanding research classification
Research outputs are grouped into categories or topics using a variety of classification systems across platforms. These systems can support strategic analysis, identify emerging fields, and help visualise research strengths and connections.
Classifications range from broad subject areas (e.g. Physics or Sociology) to highly specific topics (e.g. “deep learning in radiology”), depending on the platform and method used.
Major classification systems
Web of Science categories
-
Hierarchical structure with 254 subject categories
-
Based on journal-level classification
-
Supports traditional subject-based analysis
OpenAlex Research Topics
-
Uses machine learning to assign research topics
-
Based on concepts in article metadata and full-text
-
Offers multiple levels of granularity
-
More flexible than journal-based systems
-
Supports cross-disciplinary and emerging topic discovery
ANZSRC Fields of Research (FoR)
-
Hierarchical codes (2-, 4-, and 6-digit) developed by the Australian Bureau of Statistics and Stats NZ
-
Widely used in research funding and assessment exercises
-
Supports cross-country comparisons (particularly in Australia, NZ, and increasingly used elsewhere)
-
University of Manchester outputs are classified using ANZSRC in Altmetric Explorer and SciVal platforms
SciVal Topics and Topic Clusters
SciVal (Elsevier’s analytics platform) uses citation patterns and machine learning to create a dynamic topic-based classification system.
SciVal Topics (~94,000)
-
Smallest unit of analysis
-
Defined using citation patterns and natural language processing
-
Continuously updated to reflect emerging research areas
-
Each topic includes a group of papers sharing closely related themes
Topic Clusters (~1,500)
-
~1,500 clusters of related topics
-
Offers a higher-level view of broader research areas
-
Useful for strategic decision-making or interdisciplinary mapping
- Read Elsevier's guide to Topics and Topic Clusters
Additional Features in SciVal
Topic Prominence
-
Measures a topic’s momentum using a blend of recent citations and usage metrics
-
Helps identify fast-growing or influential areas of research
Keyphrase Analysis
-
Uses Elsevier’s Fingerprint technology to extract key concepts
-
Results are visualised (e.g. via word clouds) for easier interpretation
-
Can be applied at the topic, author, or institutional level
How these tools can help
Strategic planning
-
Identify areas of research growth
-
Find underexplored or emerging themes
-
Inform internal funding, recruitment, or REF strategies
Collaboration mapping
-
Find collaborators across disciplines
-
Spot institutional research strengths
-
Discover shared interests across Schools or Faculties
For example, we have used SciVal Topic Clusters to map research in Artificial Intelligence and Health Equity, identifying previously unconnected researchers working on related themes.
Important limitations to consider
While these systems are valuable tools, there are important limitations:
-
Methodological differences: Each platform uses its own algorithms and data sources. The same paper may be classified differently across systems.
-
Machine learning errors: Automated classifications (e.g. in OpenAlex or SciVal) can introduce inaccuracies, especially for interdisciplinary work.
-
Journal-based bias: Systems like Web of Science may misclassify papers published in newer or multidisciplinary journals.
-
Varying granularity: Some topics are too narrow, broad, or overlapping to support fine-grained analysis.
-
Access and update cycles: Not all platforms are openly available, and update frequencies vary.
These tools should be seen as guidance aids, not definitive labels. Classification data is best used in combination with qualitative insights and local knowledge.
Need support?
The Research Indicators team can help you navigate these classification systems, generate tailored reports, and interpret topic analyses responsibly.
Contact us to find out more.