Skip to content

Question Types and Preparation Focus

High-level guide for the agent (or any AI) that generates and manages practice questions for the NEC Electronics, Communication and Information Engineering Registration Examination. Pair this with:

  • strategy.md — why (exam analysis, weighting, priorities)
  • chapqsn/FORMAT.md — the exact YAML schema every question must follow
  • syllab.md — module/subchapter codes to tag against

The exam is recall- and concept-heavy MCQ, ~10 marks per module, almost no multi-step math. Generated questions must mirror that reality.

1. Question types to generate

Generate a realistic mix of these types (roughly in this priority order):

  1. Definition / concept recall — "What is X?" / "X is best defined as…" e.g. peak factor, quiescent point, weak entity, universal Turing machine, erlang.
  2. Standard value / unit — fixed numbers, units, constants. e.g. IPv6 = 128 bits, HTTPS = 443, Si diode ≈ 0.7 V, η₀ ≈ 377 Ω, A3 = 297×420 mm, Nepal AC = 50 Hz.
  3. True / False / "which is correct" — including NOT / EXCEPT / FALSE phrasing. Make exactly one option unambiguously correct; the trap is in careful reading.
  4. Confused-pair discrimination — force a choice between two similar concepts. e.g. Mealy vs Moore, Thévenin vs Norton, deadlock avoidance vs detection, RISC vs CISC, DFA vs NFA, resource levelling vs smoothing, weber vs tesla, TCP vs UDP.
  5. Application / scenario — "Which technique/device/protocol for situation Y?" e.g. which flow control for a large bandwidth-delay product; which I/O method for large blocks.
  6. Light numerical (single-step only) — one formula, no chains. e.g. RMS from peak, star↔delta, 2's complement, subnet mask. Keep math trivial.
  7. Classification / odd-one-out — "Which of the following is NOT a …".
  8. Nepal-specific & professional — NEC Act/registration, NEA, ethics, engineering economics terms.

Avoid: multi-step derivations, ambiguous/opinion questions, "all/none of the above" as filler, trick questions with two defensible answers.

2. Question format (per type)

  • Default type: mcq (single correct) — the exam standard.
  • Use type: mcq with multiple correctOptions only for genuine "select all that apply".
  • Use type: numerical only for fill-in-the-blank single-value math.
  • Always 1 correct + 3 plausible distractors (4 options total) unless the topic demands otherwise.
  • Follow chapqsn/FORMAT.md exactly (id / type / question / correctOptions / wrongOptions / explanation).
  • Math in LaTeX ($...$ / $$...$$). Every question should carry a short explanation.

3. Distraction (wrong-option) rules

  • Distractors must be plausible — common misconceptions, the "confused pair" partner, adjacent standard values (e.g. 0.3 V vs 0.7 V), or the reverse definition.
  • No obviously silly options, no joke answers.
  • Keep all options similar in length/style so the answer isn't guessable from formatting.
  • Exactly one defensible correct answer.

4. Coverage & distribution (management)

  • Mirror the exam: ~10 marks per module. Keep all 10 modules represented; never over-fill one.
  • Cover every subchapter in syllab.md — aim for a minimum number of questions per subchapter, then deepen weak areas.
  • Difficulty mix per subchapter: ~50% easy (direct recall), ~40% medium (discrimination/application), ~10% hard (edge cases). The real exam skews easy/medium.
  • Bias volume toward high-yield types: standard values, definitions, confused pairs.

5. Where questions live

  • Chapter bank → chapqsn/<CODE>-<SHORT_NAME>/<SUBCHAPTER>/<id>.yaml — one question per file, id = <subchapterCode>-<NNN> (e.g. AExE0101-001). Used for topic drilling.
  • Mock sets → mdlqs/setN.yaml — flat 100-question papers, id = setN-001…100, mixed modules, ~10 per module, mirroring the real paper. Used for full timed practice.
  • Examples/templates → mdlqs/examples.yaml — format reference only, not scored.

6. Quality & accuracy rules

  • Correctness first. Only assert an answer you're confident is right; if a source is dubious, flag it in the file (comment) and in the hand-off note. Never silently guess a key.
  • No duplicates — check existing ids and near-identical stems before adding.
  • Stable ids — never renumber or reuse an existing id.
  • Validate generated YAML (python3 -c "import yaml,glob;[yaml.safe_load(open(f)) for f in glob.glob('**/*.yaml',recursive=True)]").
  • Keep option text ASCII/LaTeX (no raw Unicode math symbols — convert ∇→\nabla, Ω→\Omega, etc.).

7. Management workflow

  1. Diagnose weak modules/subchapters (mock scores).
  2. Generate targeted questions into the relevant chapqsn/ subchapters (weak areas first).
  3. Periodically assemble a fresh mdlqs/setN.yaml mock (10/module) for timed practice.
  4. Track coverage: which subchapters still lack enough questions.
  5. Re-verify any flagged/uncertain answers before they enter a scored set.