langAz: Learn 4 Languages Through Life

langaz means language. Az means little in Turkish. 4 is little indeed in front of the world's richness of languages.

LangAz is a topic-first, multilingual learning system that teaches Spanish 🇪🇸, Portuguese 🇧🇷, French 🇫🇷, and Italian 🇮🇹 through real-life examples, not traditional grammar rules. Instead of learning "A1 to B2" or "subjunctive mood," you explore how to drive safely in rain, eat for digestion, or do the dishes without breaking anything — and learn everyday language along the way.

Built with free, open AI (LLMs like GPT-4o-mini), the app prioritizes human-centered design, relatable context, and science-based awareness — without ever calling it "physics" or "biology".

🚀 Why This Project?

We thank Michael (Mikhail) Petrunin who did a good job already. This project is here to add on to what he has started and fill in gaps for this time and the future.

As noted by language learning enthusiasts, linguist Michael Petrunin’s comparative grammar guides use tri‑ or multi‑lingual presentation (e.g. Spanish, Portuguese, French together). They are praised for structuring side‑by‑side comparisons and encouraging learners to notice differences and false friends.

Mikhail's part:

• Presents side-by-side grammar and vocabulary comparisons, a useful reference

• Advocates simultaneous study of all four languages

His work is reference-centric, not lesson- or flow-based.

Where We Can Improve Petrunin’s Comparative Grammar of Spanish, Portuguese, Italian & French

What we are to bring in:

• Personalization: no ability to choose topics freely.

• Dynamic, contextual examples tied to everyday science, math, life scenarios.

• Interactive, multimodal experience: no audio, no quizzes, no spaced repetition.

Our gap to fill: Create a learner-driven, topic-first, multi-Romance AI-powered system with scientific and real-world contextualization—not level-bound. Generate tri-lingual lessons, audio, practice, and quizzes, all anchored in everyday science, health, ethics, physics, etc.

This project:

• Demonstrates cutting-edge use of GenAI and agentic workflows.
• Trains multilingual reasoning through sentence comparison.
• Teaches scientific and well-being concepts through life-based scenarios.
• Uses AI to personalize learning top-down, not bottom-up.
• Is free, fully open source, and future-extensible (voice/video/agents later).

🔍 What You Learn (Without Even Noticing)

Instead of memorizing verb tables, you'll absorb:

• How different Romance languages express the same everyday experience.
• How grammar differs (reflexives, tense, sentence structure) across languages.
• Subtle patterns that unlock fluency.
• Concepts of energy, momentum, digestion, habits — all through story-like sentences.

🧱 Structure

Each lesson includes:

• A real-life topic: e.g., "Why trucks stop slower than bikes".
• Short 1-2 line sentences in ES, PT, FR, IT.
• Optional explanations and grammar/word comparisons.
• Progressive deep dives: learners can expand on what interests them.

Lessons are:

• 📌 Topic-first (not level-based)
• 🧩 Modular (choose what to expand: verbs, pronunciation, grammar, etc.)
• 📖 Text-only (for now)
• 💡 Built for future AI agents, TTS, and SRS integration

🛠 Technologies Used

• GPT-4o-mini and open LLMs for multilingual generation
• Python for content generation and packaging
• (Optional frontend) Streamlit or plain markdown
• GitHub Pages (optional future deployment)
• Markdown + JSON content format
• Fully extensible to Hugging Face Spaces or mobile

🧪 Example Output

Topic: Driving safely on wet roads

Spanish: El coche tardó más en frenar porque la carretera estaba mojada.
Portuguese: O carro demorou mais para parar porque a estrada estava molhada.
French: La voiture a mis plus de temps à s’arrêter parce que la route était mouillée.
Italian: L’auto ha impiegato più tempo a fermarsi perché la strada era bagnata.
English: The car took longer to stop because the road was wet.

Click to expand:

• 🔤 Compare verb structure (tardar, demorou, a mis, ha impiegato)
• 🧠 Concept: heavier objects take longer to stop (momentum)
• 📣 Pronounce “mojada” (mo-ha-da), “molhada” (mol-ya-da), “bagnata” (ban-ya-ta)

📂 File Structure

.
├── content/
│   ├── driving/
│   ├── digestion/
│   ├── cleaning/
│   └── ... (more topics)
├── generator/
│   └── generate_lessons.py
├── ui/
│   └── (optional Streamlit app)
├── README.md
├── ROADMAP.md
└── LICENSE

🧑‍💻 Authors & Credits

Built by Yashine Hazmatally Goolam Hossen (my family calls me "Az", hence the name langAz - also meaning language in Mauritian Creole), integrating insights from comparative grammar, polyglot workflows, and practical neuroscience. Inspired by the start of the Portuguese movie, "Papa ne comprend pas" being watched on Netflix (July 21, 2025) while coding this.

Some Topics To Be Covered

Driving and Safety (braking, awareness, heavy vehicles, rain)

Cooking and Eating (cheese and wine, digestion, food order)

House and Cleaning (dishwashing, lifting heavy things, organizing shelves)

Walking and Movement (momentum, balance, slipping, posture)

Helping Others (feeling peaceful, hormones, smiling)

Daily Habits (sleeping in dark, sipping water, listening to music)