Find the literature behind your research question
New Semantic Search in Scopus helps you explore related publications from a research question, even when the literature uses different terminology.
Introducing Semantic Search in Scopus
An engineer investigating force feedback in robotic surgery encounters work on haptics, instrument sensing and tissue interaction. The research question can be precise while its vocabulary is still unsettled. Semantic Search provides a way into that literature through the question itself.
From a question to candidate studies
Sign in to Scopus, open Semantic Search and enter a focused question. For example:
How do force-feedback settings affect tissue-handling forces during robotic surgical tasks in bench and ex-vivo models?
Enter the research question in Semantic Search to explore related publications.
Semantic Search uses vector search with language-model embeddings: numerical representations of meaning that help match a question with conceptually related document content. It retrieves indexed publication records for inspection. Meaning-based matching does not establish equivalent methods or complete coverage.
In corporate R&D, every decision brings opportunity and risk - and speed matters. Scopus Semantic Search lets researchers search in their own language, going beyond keywords to uncover relevant research they might otherwise miss. Instead of relying on AI-generated answers only, researchers stay in control, quickly evaluating evidence to reduce risks and make faster, better-informed decisions.Evgeny Kryukov
Senior Director Product Management, Scopus at Elsevier
Choose the evidence for your investigation
Review titles and abstracts, then refine the results with the available filters. For this question, a clinical paper and a bench or ex-vivo study may address different parts of the problem. Examine the task, tissue model, feedback setting and force measurement before grouping findings.
Review candidate studies and their experimental settings. The returned list still requires relevance assessment.
Where Create AI summary is enabled, select up to 20 references that form a coherent set. The overview uses those selected abstracts. Check its statements against the supporting sources and read the papers for methods, results and limitations.
An excerpt from the generated summary of selected abstracts. Check its statements against the supporting sources.
Let the research decision define the evidence set
For an antibody–drug conjugate researcher, the same workflow separates discovery from selection. Related terminology may lead to studies of plasma stability or intracellular release. Distinguish those endpoints before choosing abstracts for a summary; the relevant model and assay determine what each paper can contribute to the investigation.
An engineer exploring lithium recovery from brine faces a different comparison. Feed composition, competing ions and regeneration conditions can change how a reported selectivity result should be interpreted. Use meaning-based discovery to find candidates, then define a comparable subset for closer reading.
Frequently asked questions
It matches meaning, rather than relying only on identical wording. Review the retrieved records for relevance.
Where enabled, select up to 20 references from the results before choosing Create AI summary.
It uses the selected abstracts. Inspect the papers for full methods, findings and limitations.
No. Assess coverage, search strategy and screening; meaning-based ranking does not establish completeness.
Use it to inspect and edit an explicit Boolean strategy before running a conventional Document search.
Sign in and check your enabled features and current allowance. Follow organizational AI policies; exclude personal, confidential or sensitive information.