EduSimplify

French text simplification + CEFR analysis + text-to-speech for language learners
Built for real learners, teachers and EdTech use cases.

---

What is EduSimplify?

EduSimplify is a small NLP-powered web app that helps French learners and teachers:

• Simplify French texts using rule-based + frequency-based transformations
• Estimate CEFR level (A1–C1) of the original and simplified text
• Highlight difficult words based on frequency (wordfreq)
• Read the simplified text aloud (Text-to-Speech) with speed and voice options

It is designed as a practical, didactics-aware tool for:

• teachers preparing materials at the right level
• learners who want to understand “why this text feels hard”
• EdTech experiments in explainable, rule-based simplification

---

Main Features

• 🔍 CEFR-like difficulty analysis

  • Heuristics using sentence length, lexical frequency, and rare word proportion
  • Returns an estimated level (A1–C1) + a band (e.g. A2–B1) + explanation text

• Rule-based text simplification

  • Simplifies connectors (cependant → mais, nonobstant → malgré…)
  • Rewrites some heavy academic / administrative expressions into simpler language
  • Uses word frequency (wordfreq) to replace rare words when possible
  • Adapts behaviour depending on target level (A1, A2, B1, B2, C1) and mode (light / standard / strong)

• 🗺️ Two strategies

  • auto → choose a target level automatically based on the original text
  • target → user selects an explicit CEFR level (A1–C1) for the simplification

• 🎨 Visual lexical feedback

  • Original text is highlighted according to lexical difficulty:
    • green = frequent words
    • yellow = medium frequency
    • red = rare words (potentiellement difficiles)

• 🔊 Text-to-Speech for the simplified text

  • Uses browser’s SpeechSynthesis API
  • Speed options: slow, normal, fast
  • Voice preference: automatic, “female”, “male” (best-effort based on available voices)

---

How it works (NLP / Didactics)

Under the hood:

• spaCy (fr_core_news_sm)

  • sentence segmentation
  • POS tags and lemmas
  • used both for simplification rules and CEFR-like analysis

• wordfreq

  • word frequency scores on Zipf scale (0–7)
  • defines “easy / medium / hard” words
  • used to mark rare words and decide candidates for substitution

• Custom rules / patterns

  • multi-word expressions:
    • « il convient de noter que » → « il faut dire que »
    • « la dichotomie entre » → « la différence entre »
  • structural patterns:
    • « Il est ADJ de… » → C’est ADJ de / C’est ADJ. On doit… depending on level
    • some simple passive forms → « est fait par » for lower levels

• CEFR-like heuristic

  • counts sentences and tokens
  • average sentence length
  • ratio of rare words (“hard”)
  • maps these indicators to a rough CEFR estimate (for demo/prototype use, not for official certification)

This is intentionally transparent and rule-based, so it can be discussed with teachers and learners.

---

Demo workflow

1. Paste a French text (e.g. from news, literature, exam prep).
2. Choose:
   • Mode: light / standard / strong
   • Strategy: automatic vs specific CEFR target
3. Click “⚙️ Simplifier et analyser”
4. See:
   • CEFR box with estimated level + description
   • Original text with coloured lexical difficulty
   • Simplified version with its own CEFR estimation
   • Strategy explanation (how / why the simplifier decided)
5. Optionally click “🔊 Lire le texte simplifié” and adjust speed or voice.

---

🛠️ Tech Stack

• Backend

  • Python 3.10+
  • FastAPI
  • spaCy (fr_core_news_sm)
  • wordfreq

• Frontend

  • Vanilla HTML + CSS + JavaScript
  • Fetch API for communication with FastAPI
  • Browser Text-to-Speech (SpeechSynthesis)

---

Run the project locally

1. Clone the repository

git clone https://github.com/Conyekp2/EduSimplify.git
cd EduSimplify

2. Create and activate a virtual environment

python3 -m venv .venv
source .venv/bin/activate  # macOS / Linux
# .venv\Scripts\activate   # Windows (PowerShell / CMD)

3. Install dependencies

pip install -r requirements.txt
python -m spacy download fr_core_news_sm

4. Start the FastAPI server

uvicorn app.main:app --reload

5. Open the web interface

http://127.0.0.1:8000/static/index.html

Project structure

EduSimplify/
├─ app/
│  ├─ __init__.py
│  ├─ main.py        # FastAPI app (API + static file serving)
│  ├─ simplify.py    # Simplification pipeline (rules + frequency)
│  └─ cefr.py        # CEFR-like analysis + lexical difficulty
├─ static/
│  └─ index.html     # Frontend UI (textarea, controls, results, TTS)
├─ requirements.txt  # Python dependencies
├─ .gitignore
└─ README.md

Author

Chinedu Onyekpere Multilingual NLP practitioner & EdTech-oriented language teacher. Focus: NLP for learning, CEFR-aligned tools, explainable simplification.

GitHub: https://github.com/Conyekp2

LinkedIn: https://www.linkedin.com/in/chinedu-onyekpere-5a89912a4/