Knowledge Graph App is an open, inter-operable, interactive tool for building, visualizing, and editing scientific and semantic knowledge as a graph of propositions encoded in an open standard (RDF). It is evolving as a learning studio, enabling learners to think like a scientist. It is intended to be used by neo-literates to experts. Designed for clarity (rigor), and modularity without loosing expressiveness. it enables users to construct, inspect, and extend both binary relations (connections between concepts) and attribute-value assignments (key-value properties), all using a user-friendly visual interface built with React and Blockly with an extensible graph-based backend.
Why? Because scientific knowledge is best represented as composable propositions—not just entities and links, but also statements about those entities, their attributes, and their relationships. This tool supports building knowledge that is both human-readable and machine-actionable.
-
Modular Propositions: Every piece of knowledge is a proposition—either a relation (subject–predicate–object) or an attribute assignment (subject–attribute–value).
-
Node-Centric Composition: Clicking a node presents a full, readable stack of everything known about it: all its relations (grouped by type), all its attribute-value pairs, and all quantifiers or qualifiers: "All - human beings - belong to → living creatures"; "Isaac Newton - date of birth → 1643-01-04.
-
Grouped Relations: If a node has several outgoing edges of the same relation type, they’re grouped visually for clarity—e.g., “Isaac Newton - member of → Human being, Scientist, Mathematician” appears as one grouped block.
-
Custom Attribute Blocks: Attribute values are input and displayed using fields that match their data type (number, boolean, enum, date, etc.), ensuring semantic correctness and clean data export.
-
Backend Sync: All propositions are stored and retrieved via a Python (Flask/SQLite) backend, so what you see in the UI is what is stored in the knowledge base.
-
Block-Based UI: Built on Blockly, every proposition is an editable, draggable block—no clutter, no hidden magic.
- Frontend: React, Blockly, modern JavaScript
- Backend: Python (Flask), SQLite
- APIs: REST endpoints for nodes, relations, attributes, and node-attribute assignments
-
Add Nodes: Create concepts or entities/individuals (e.g., “Water”, “Cell”, “Einstein”).
-
Add Relations: Link nodes via named relations (“is a”, “part of”, “member of”, etc.), with group support for one-to-many relations.
-
Add Attributes: Assign data-rich properties to nodes (e.g., “boiling point: 100°C”), with full type and value validation.
-
Compose Node Profiles: Clicking a node injects a compose block with:
- The node
- All its outgoing relations, grouped by type
- All its attribute-value blocks
-
Visual Edit and Export: Drag, reorder, or edit blocks. Export the entire proposition stack for backend updates or downstream tools.
A knowledge profile for Water might look like:
-
[Water]
- is a → [Compound]
- part of of → [Hydrosphere], [Hydrologic Cycle]
- boiling point: 100* *
-
{at room temperature} [Water]
- state: liquid
A summary of the node is parsed and structured according to POS (parts of speech), verb phrases, prepositions, logical connectives and so on, helping learners to bring meaning to the foreground.
Words acquire meaning from its relationship with others, and not by themselves, basing on a sound network theory of meaning. This app is inspired by the needs of scientific reasoning, education, and semantic interoperability—where knowledge is always a web of claims, not just disconnected nodes or edges. We build for clarity (rigor), modularity, and extensibility.
The App is not ready for end users as yet.
-
Clone the repo:
git clone https://github.com/gnowgi/knowledge-graph-app.git cd knowledge-graph-app -
Install dependencies and run:
- Backend:
cd backend && pip install -r requirements.txt && python app.py - Frontend:
cd frontend && npm install && npm start
- Backend:
-
Open http://localhost:8080 in your browser.
-
Full proposition extraction/export in a controlled natural language (CNL)
-
Saving visualizations in SVG and other image formats.
-
A quantitative rubric to compute knowledge base profile
-
Support for
-
Process Modeling using Petri Nets
-
Procedural Knowledge (Scratch/Logo)
-
Concept Mapping
-
Logic:
- first order and second order calculus
-
model based reasoning
-
data analysis and visualization
-
collaboration
-
*(Coming soon) *
AGPL v3
For issues, suggestions, or contributions, please open an issue or submit a pull request