Explainable AI Toolkit for Scientific Data

This project demonstrates a lightweight framework for:

• Knowledge Graph construction
• Symbolic reasoning
• Causal explanation of events
• Natural language summarization
• A REST API for explainability

It aims to show skills relevant to research internships in AI, Explainable AI (XAI), Knowledge Graphs, and symbolic reasoning (Inria BOREAL, Coconut, EDF LLM, TALN Antilles...).

---

Concepts covered

Knowledge Graphs

Graph integration of heterogeneous data (sensors, events, causes). Using Python + NetworkX.

Symbolic AI & Reasoning

Shortest-path explanation of causal chains. Root cause analysis for scientific events.

Natural Language Processing (NLP)

Simple text generation for event summarization.

API Development

FastAPI server exposing:

• /neighbors/{id}
• /explain/{event_id}
• /summary/{event_id}

---

Project Architecture

explainable-ai-toolkit/ ├── data/ ├── kg/ ├── reasoning/ ├── nlp/ └── api/

How to run

pip install -r requirements.txt uvicorn api.api:app --reload

Copier le code http://localhost:8000/explain/Event2

Purpose

This mini-project demonstrates abilities in:

• Data modeling & knowledge representation
• Reasoning & explainability
• Python architecture
• NLP
• Scientific data processing
• Research-oriented problem solving

It fits 100% the expectations of research teams working on Explainable AI, Knowledge Graphs, Reasoning, NLP, or XAI.