Graph

Graph is ResearchArk's research intelligence platform, accessible at researchark.eu/graph. It combines an AI-powered answering engine with structured data browsing across the European and international research landscape. Backed by a Neo4j knowledge graph, PostgreSQL database, and Weaviate vector store, Graph provides both conversational AI access and traditional browsing interfaces for navigating the relationships between projects, organizations, funding schemes, and research outputs.

Navigation

Graph is organized into nine sections, grouped by data scope:

Data Sources

Graph draws from multiple authoritative sources to build a cross-linked picture of the research ecosystem:

• CORDIS (Community Research and Development Information Service) -- The primary source, covering EU-funded research projects from Framework Programme 1 through Horizon Europe. Includes project metadata, participant lists, budgets, topic classifications, publications, and deliverables.
• KOHESIO -- EU Cohesion Policy project data with over 2 million operations, covering regional development and structural fund investments. Includes operational programmes, fund distribution, beneficiaries, and regional breakdowns.
• INTERREG -- Cross-border, transnational, and interregional cooperation programme data. Browsable by projects, countries, themes, programmes, and partners.
• ERA-LEARN -- European Partnership initiatives including joint programming, co-funded research activities, calls, and participating organizations.
• International sources -- Selected datasets from non-EU funding bodies across Spain, Ireland, Sweden, the UK, and the US. Organized by countries, funders, organizations, programs, grants, people, and outputs.

These datasets are cross-linked within the graph, enabling you to trace connections that span programme boundaries.

Knowledge Graph Architecture

The underlying data infrastructure combines four complementary systems:

• Neo4j -- Graph database storing entity relationships. Node types include Project, Organization, FrameworkProgramme, Topic, EuroSciVoc, LegalBasis, Country, FundingScheme, Publication, Deliverable, KohesioOperation, KohesioBeneficiary, and others.
• PostgreSQL -- Relational database for structured queries, filtering, aggregation, and keyword search.
• Weaviate -- Vector database using 768-dimensional embeddings from EmbeddingGemma-300M for semantic search across projects and other entities.
• LLM Service -- AI answering layer powered by Gemini models (2.5 Flash Lite through 3 Pro), generating streaming responses with inline visualization directives.

When you query the Explorer, all four systems are searched in parallel. The AI synthesizes results from keyword matches, vector similarity, graph traversals, and SQL aggregations into a single coherent answer.

Key Capabilities

• AI-powered research answers -- Ask natural-language questions and receive streaming responses with source indicators, inline charts, tables, and statistics.
• Structured data browsing -- Browse projects, organizations, topics, and funding schemes with filters, search, sorting, and pagination.
• Stakeholder dashboards -- Purpose-built analytics views for researchers, institutions, businesses, and funders, each surfacing the metrics most relevant to that audience.
• Cross-dataset exploration -- Navigate between EU CORDIS, Cohesion Policy, and international funding data through a unified interface.
• Progressive loading -- Dashboard components load independently with skeleton screens, so fast data sources appear immediately while slower queries complete in the background.

Explore Graph