Proofreading under penalty: which eight doubtful spots to verify
After a 10-minute Hindi typing test, a candidate has 3 minutes of proofreading time. The software does not underline errors. The marking scheme is: full error = 2 penalty points, half error = 1, quarter error = 0.5; a wrongly corrected word is treated as a new full error. The candidate has flagged 12 doubtful spots: five are matra pairs that change the word (कि/की, तु/तू, है/हैं), three are punctuation or spacing (। versus ., space before comma), two are श/ष or ब/व substitutions, and two are uncertain conjuncts with reph/halant. She has time to verify only eight. She is tempted to correct all 12 quickly. Her current typed text otherwise matches the passage. She needs to decide which to check first and whether to touch the uncertain conjuncts.
1. Rank the 12 doubtful spots by expected value. Which eight should she verify first?
Read the worked answer
Use expected value, not visibility. A full error is worth 2 points, a half error 1, a quarter error 0.5, and a wrong correction costs a new full error. The five matra pairs come first: they are full errors, they change meaning, and they can be checked quickly by reading the word. Next come the two श/ष or ब/व substitutions: also full errors, but they require a word-meaning check. Then one punctuation or spacing group, because those are fast and carry half or quarter penalties.
That gives eight checks: five matra pairs, two consonant substitutions, and one punctuation group. The two uncertain conjuncts should be left alone unless time remains, because a wrong conjunct correction creates a new full error. The remaining punctuation groups are lower priority than the full-error matras and consonants. If a matra check is genuinely uncertain after 20 seconds, skip it rather than guessing.
2. Show the cost of a wrong correction and explain why uncertain conjuncts are poor targets.
Read the worked answer
If she changes a correct word, the marking scheme treats it as a new full error, costing 2 points. If she leaves a real full error, she loses 2 points. Suppose a doubtful conjunct has a 50% chance of being wrong and a 50% chance of being correct. Fixing it has expected gain = 0.5 × 2 − 0.5 × 2 = 0. There is no expected benefit, but there is added risk of breaking a correct conjunct. If the chance of error is 80%, expected gain = 0.8 × 2 − 0.2 × 2 = 1.2 points, which is worth doing. If it is 30%, expected gain = 0.3 × 2 − 0.7 × 2 = −0.8, so skip.
Uncertain reph/halant conjuncts are especially poor targets because the visual difference can be subtle and a quick correction may produce a form that is not a valid word. Certain matra pairs are better because the correct word is usually obvious from the sentence.
3. Give a minute-by-minute proofreading plan for the 3 minutes.
Read the worked answer
Minute 1: verify the five matra pairs using the sentence context. Read the whole word, not just the matra. Minute 2: verify the two श/ष or ब/व substitutions and one punctuation group; fix only high-confidence cases. Minute 3: do a fast scan for spacing before punctuation, double spaces and sentence-ending marks, then stop editing. Do not reread the entire passage from the beginning; that wastes time and increases the chance of unnecessary changes.
A practical check is 20–25 seconds per doubtful spot. If a spot cannot be resolved in that time, leave it. She should also avoid retyping whole words when a single matra or punctuation mark can be replaced. The goal is to maximize expected score, not to eliminate every visible doubt.