Research AI Landscape

knowledge-bases

3 min read

Knowledge Bases

Overview

Covers tools for building structured, queryable, and AI-ready knowledge stores over research information. This includes personal knowledge management (PKM) tools, collaborative knowledge graphs, semantic triplestores, vector databases, and RAG (retrieval-augmented generation) orchestration frameworks.

Disambiguation:

  • knowledge-bases — organising, connecting, and retrieving knowledge; powering AI over research content; semantic infrastructure. Tools: Obsidian, Wikibase, OpenAlex, GraphDB, Neo4j, Qdrant, Weaviate, LlamaIndex, LangChain, Azure AI Search, NotebookLM, Confluence.
  • research-data-management — storage, preservation, and compliance for research outputs; FAIR data; repositories; DMPs; ELNs; CRIS. Tools: DMPonline, Figshare, Zenodo, RSpace, LabArchives, Symplectic Elements.
  • data-collection — primary data capture; surveys; participant recruitment. Tools: Qualtrics, Prolific.

The boundary between knowledge-bases and research-data-management is: if the primary question is “how do I find and use this information?” it belongs here; if it is “how do I store and preserve this output compliantly?” it belongs in RDM. Tools with a foot in both (e.g. a CRIS that also powers AI search) should be listed under both areas.

Tools in this area

ToolVendorMaturityStatusCost (GBP/yr)
obsidianDynalist Inc.production-readytrial0
notebooklmGoogleproduction-readyadopt0
ii-commonsII-Commons (open source)experimentalassessunknown

Further tools to be added as notes are drafted — see Research knowledge infrastructure report for the full landscape (Wikibase, OpenAlex, GraphDB, Neo4j, Qdrant, Weaviate, LlamaIndex, LangChain, Azure AI Search, Confluence).

Landscape notes

PKM as research infrastructure: Personal knowledge management tools (obsidian, notebooklm) are the most immediately useful in this area — low friction, free, and already available at Sheffield. notebooklm is included in Sheffield’s Google Workspace licence.

AI grounding stack: The report identifies a clear recommended open-source stack for institution-level AI knowledge infrastructure: Azure AI Search (leveraging existing Microsoft agreements) for managed RAG with UK data residency; Qdrant or Weaviate (self-hosted) for sensitive data requiring full sovereignty; LlamaIndex + LangChain for orchestration with source provenance. No Russell Group university has yet deployed this at institutional scale — a first-mover opportunity.

Knowledge graphs for humanities: Wikibase (the platform behind Wikidata) has proven applicability for digital humanities special collections and library catalogues. Zero licensing cost, SPARQL-queryable, and well-aligned with FAIR principles. UCL’s Sloane Lab (British Museum/Natural History Museum) used GraphDB + metaphactory for a UKRI-funded project — the only confirmed Russell Group knowledge graph deployment.

GDPR risk profile: Significant variance within this segment. EU-headquartered, self-hostable tools (Qdrant, Weaviate, GraphDB, Wikibase) carry low risk. US SaaS-only tools (Pinecone) carry high risk and should be avoided for research data. See research-knowledge-base-infrastructure report for tool-by-tool GDPR assessment.

Emerging: Graphwise (Ontotext + Semantic Web Company merger, Feb 2025) could become the leading EU-based knowledge graph + AI platform. Monitor through mid-2027.

Key intelligence

research-data-management · literature-review · data-collection · ideation-and-grant-writing

Sprint history

  • 2026-03-05: Stub created — split from research-data-management following incorporation of knowledge infrastructure report

← Back to main landscape

All capability areas: Ideation and Grant Writing · Literature Review · Hypothesis Generation · Data Collection · Automatic Transcription · Autonomous Laboratories · Data Analysis · Scholarly Writing · Qualitative Coding · Knowledge Bases · Research Data Management

Graph View

Backlinks