AI Math MCQ Generator
Budget: ₹1,500 – ₹12,500 INR
I need a small AI-powered script that reads ordinary text documents and automatically turns their mathematical content into hard-level multiple-choice questions. The input will always be plain .txt files; no PDFs or web scraping is required.
Here’s what I expect: the program (Python is preferred, with open-source NLP or LLM libraries such as spaCy, NLTK, or transformers) ingests a text file, detects mathematical statements, definitions, theorems, or worked examples, and formulates challenging MCQs based on them. Each item must include:
• a clear stem that tests understanding rather than recall
• one correct answer and three plausible distractors
• an answer key and brief explanation
Hard difficulty means the questions should demand multi-step reasoning, not simple plug-and-chug arithmetic. Wherever the source text lacks enough depth, the algorithm should synthesise context so the question still feels self-contained.
Acceptance criteria
• Runs locally from the command line: python generate_mcq.py input.txt output.json
• Generates at least 10 MCQs per 2,000 words of source text.
• Output JSON validates against a simple schema I’ll share after award.
• Code is clean, commented, and accompanied by a short README explaining setup and model choices.
If you have prior work in question generation or adaptive learning systems, let me see a sample when you bid.
Here’s what I expect: the program (Python is preferred, with open-source NLP or LLM libraries such as spaCy, NLTK, or transformers) ingests a text file, detects mathematical statements, definitions, theorems, or worked examples, and formulates challenging MCQs based on them. Each item must include:
• a clear stem that tests understanding rather than recall
• one correct answer and three plausible distractors
• an answer key and brief explanation
Hard difficulty means the questions should demand multi-step reasoning, not simple plug-and-chug arithmetic. Wherever the source text lacks enough depth, the algorithm should synthesise context so the question still feels self-contained.
Acceptance criteria
• Runs locally from the command line: python generate_mcq.py input.txt output.json
• Generates at least 10 MCQs per 2,000 words of source text.
• Output JSON validates against a simple schema I’ll share after award.
• Code is clean, commented, and accompanied by a short README explaining setup and model choices.
If you have prior work in question generation or adaptive learning systems, let me see a sample when you bid.