Generate Clarification Questions with Simple NLP

[salazarsebas]

❓ Generate Clarification Questions with Simple NLP 🤔

📝 Description

Create a backend function to generate 2-5 natural-language clarification questions based on the parsed project idea. This uses simple keyword matching (mocking LLM behavior) to identify gaps, like data needs or Stellar specifics, for a non-technical chat flow.

🎯 Objective

Implement a standalone function in /backend to produce targeted questions from idea keywords, enhancing the interactive clarification step.

✅ Requirements

• In /backend, create a new file clarificationGenerator.js.
• Implement a function generateQuestions(ideaText) that extracts keywords (e.g., regex for "remesas", "datos") and maps to predefined question templates (e.g., if "Stellar" mentioned, ask "Qué datos sensibles necesitas?").
• Return an array of 2-5 questions as strings, with visual-friendly phrasing (e.g., including emojis).
• Handle edge cases: If no keywords, return generic questions.
• Export the function for use in routes like /chat/init.
• Commit changes to the Git repository with a message like "Implemented clarification question generator with simple NLP".

🏆 Expected Outcomes

• clarificationGenerator.js is created and the function returns relevant questions (e.g., input "App remesas Stellar" → ["¿Qué volumen de transacciones esperas? 💸", "¿Datos off-chain como emails? 📧"]).
• Questions are natural and non-technical, limited to 2-5 per call.
• Function is testable independently via Node REPL.
• Changes committed, and integrates with existing parsing logic.

🔗 References

• JavaScript Regex for Keyword Extraction
• Simple NLP in JS (optional library; use if installed, else pure JS)

📋 Notes

• Keep it mock-simple: No real LLM; use if/else or switch on keywords for now.
• Test with sample inputs: Log questions to console and verify variety.
• Later, replace with full LLM integration for advanced parsing.

[macharra]

Hi, I’d like to take this issue. I have experience with JavaScript, regex, and simple NLP techniques, and I can implement the generateQuestions(ideaText) function exactly as described.

My approach:

Create clarificationGenerator.js in /backend.

Use regex/keyword matching to detect terms (e.g., “Stellar”, “remesas”, “datos”).

Map keywords to predefined, emoji-friendly clarification question templates.

Handle edge cases by returning 2–5 generic questions when no keywords match.

Make the function testable independently in Node REPL and export it for routes like /chat/init.

I’ll commit with a clear message and ensure code readability so it can later be swapped with an advanced LLM implementation.

I’m confident I can deliver a working version quickly and iterate with feedback.

[salazarsebas]

Just to clarify: The changes required for this issue are in the stellar-idea-forge folder. @macharra

[salazarsebas]

Hey @macharra , what are the updates on this issue?