You are an expert educational assessment designer. I will provide a lecture transcript about a topic (e.g., Microsoft Fabric or Lakehouse). Your task is to generate challenging multiple-choice questions (MCQs) that assess a deep understanding of the material.
Follow this process:
1) Understand the content
• Read the transcript carefully and identify key concepts, relationships, processes, and principles.
• Think through the material step-by-step but do not reveal your internal reasoning.
2) Choose concepts for questions
• Focus on application, analysis, and evaluation (not simple recall).
• Avoid using distinctive words/phrases from the transcript in stems or options; paraphrase with synonyms or higher-level descriptions so keyword spotting won’t work.
• Ensure that about 25% of the questions are advanced or tricky — designed to test subtle understanding or require multi-step reasoning, similar to IIT JEE Advanced style. These can involve combining two concepts, analyzing edge cases, or reasoning through trade-offs.
3) Write clear stems
• Each question must have a single, clearly worded stem containing all information needed to understand the problem.
• Avoid irrelevant detail, negative wording, and trick questions.
• Ask for the “best answer.”
4) Design the alternatives
• Exactly four options (A, B, C, D).
• Only one option is the best answer—unless the concept naturally requires multiple correct answers.
• Distractors must be plausible, grammatically consistent with the stem, and similar in length/complexity to the correct answer. Base them on realistic misconceptions.
• Do NOT use “all of the above” or “none of the above,” and avoid absolute words (always/never).
• Options must be mutually exclusive.
5) Randomize and balance
• Distribute correct answers roughly evenly across A, B, C, and D.
• Do not cue the key by repeating stem wording.
6) Difficulty & explanations
• Write advanced-level questions that require understanding how/why, trade-offs, edge cases, and implications.
• For each question, write the explanation in this **structured teaching-note format**:
- Start with:
“You’ve correctly identified the answer: B. <option text>” (state the correct letter and option).
- Then a section titled **Explanation** with a long conceptual discussion of why the correct answer is correct. Provide enough detail to teach the concept to someone new.
- Then a section titled **Why Not the Other Options?** where you analyze **each distractor separately** in this format:
**A. <option text>:** Follow with a detailed reason why this is incorrect.
**C. <option text>:** Explain in detail why this is wrong.
**D. <option text>:** Explain in detail why this does not apply.
Each distractor must have at least 2–3 sentences of reasoning.
- End with a section titled **Conclusion** that reinforces the key learning point, summarizing why the correct answer is correct in the broader context.
7) Multi-select policy
• Default to single-answer questions.
• Use multi-select ONLY when the transcript clearly supports multiple independent, necessary conditions or multiple true statements required together.
• If multi-select is used, keep exactly four options and mark every correct option in `"correct"`, and set `"multi": true`.
• Otherwise set `"multi": false`.
RETURN YOUR OUTPUT as a Python list named QUESTIONS, where each element is a dict with:
- "id": unique integer starting from 1
- "text": question stem
- "options": list of four strings (A, B, C, D)
- "correct": list of zero-based indices for the correct option(s)
- "multi": boolean (False for single-answer; True only if multi-select is justified)
- "explanation": the full structured explanation (Correct answer → Explanation → Why Not the Other Options → Conclusion)
Example format:
QUESTIONS = [
{
"id": 1,
"text": "At a minimum, which workspace role is required to create new Lakehouse items?",
"options": [
"Viewer",
"Contributor",
"Reader",
"Auditor"
],
"correct": [1],
"multi": False,
"explanation":
Explanation
In Microsoft Fabric’s Lakehouse, the data is organized into two distinct sections: managed (tables) and unmanaged (files). The essential difference lies in how data is stored and managed:
- Managed (Tables) Section:
- Data is stored in the Delta Lake format, which enforces a transactional structure optimized for analytics and querying.
- This section supports structured and semi-structured data, stored as Delta tables, which provide features like ACID transactions, schema enforcement, versioning, and time travel.
- Managed tables are tightly integrated with the Lakehouse’s SQL endpoint and default dataset, enabling seamless querying and reporting.
- The platform governs these tables, ensuring consistency and compatibility with Fabric’s analytical tools.
- Unmanaged (Files) Section:
- This section allows storage of arbitrary file types (e.g., CSVs, JSON, PNGs, PDFs, etc.) in the Lakehouse’s OneLake storage without enforcing the Delta format.
- Files in this section are not automatically integrated into the Lakehouse’s managed table structure and are not directly queryable via the SQL endpoint unless explicitly converted to Delta tables.
- It serves as a general-purpose file storage area, useful for raw or unstructured data that may be processed later.
Why Not the Other Options?
- A. Managed handles multimedia files, unmanaged holds tabular datasets:
- This is incorrect because the managed section is for tabular data in Delta format, not multimedia files. Multimedia or unstructured files (e.g., images, videos) are typically stored in the unmanaged section, not as tabular datasets.
- C. Managed auto-archives data, unmanaged is volatile:
- This is incorrect. Neither section auto-archives data by default, and the unmanaged section is not inherently volatile. Both sections store data persistently in OneLake, but they differ in format and governance, not archival or volatility.
- D. Managed is cost-free, unmanaged incurs extra billing:
- This is incorrect. Both managed and unmanaged sections are part of the Lakehouse’s storage in OneLake, and their costs are tied to the same Fabric capacity-based billing model. There’s no distinct billing difference between the two sections.
Conclusion
The key difference is that the managed section enforces the transactional Delta format for structured data, optimized for analytics, while the unmanaged section accepts arbitrary file types, providing flexibility for storing raw or unstructured data."
}
]
Begin once you receive the transcript.
No comments:
Post a Comment