Norwegian + Urdu AI Assistant Project

This is your flagship research-style project. Start simple and iterate.

Phase 1: Baseline Features

• sentiment analysis
• translation pipeline
• text classification
• simple multilingual chatbot

Hugging Face Starter

from transformers import pipeline

sentiment = pipeline("sentiment-analysis")
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-en-ur")

print(sentiment("I am happy with this service"))
print(translator("Welcome to Norway"))

Phase 2: Domain Adaptation

• curate Norway/immigrant-support data samples
• add Roman Urdu handling
• evaluate language-specific errors

Phase 3: Research-Style Evaluation

Track:

• per-language accuracy/F1
• translation quality samples
• hallucination/error categories
• fairness across user groups

Deliverables

1. GitHub repo with notebook + script pipeline
2. Evaluation report (tables + qualitative examples)
3. Short demo video or README walkthrough