The actual look of my proofreading notes before a chapter of a Marathi novel goes to print

Around 11 p.m. last night, I was about to close the spreadsheet on my desk when one cell caught my eye. "3rd revision, sentence 127, fidelity score 6/10, readability score 4/10." Nine sentences in the chapter still had low scores on both counts. The deadline was two days away. I wanted to just turn off the monitor, but I brewed another cup of coffee.

From the outside, translation looks fairly simple. You read the source text, transfer the meaning, polish it, and send it off. In reality, by the time a single chapter reaches a printable state, at least four or five versions of the same file have piled up on my desktop. Today, I want to show that process without sugarcoating anything.

The Spreadsheet Is the Map

I translate Marathi novels into English. Marathi uses the same Devanagari script as Hindi, but its sentence structures and idiomatic expressions are quite different. In novels especially, colloquialisms, regional dialects, and generational speech patterns vary so widely that "an accurate translation" never exists as a single right answer.

So after I first translate a chapter, I number every sentence and organise them in a spreadsheet. The columns look like this:

• Sentence number
• Source text (Marathi)
• First-draft translation (English)
• Fidelity score (1-10): how well the original meaning and nuance have been preserved
• Readability score (1-10): how natural it reads for an English-language reader
• Notes: why I gave this score, what alternative phrasings I considered

I score all of this myself. It's entirely subjective, and the scores can fluctuate depending on how I'm feeling that day. Still, it's better than just going on instinct and thinking "this sentence feels a bit off" before moving on. When things are recorded as numbers, it's immediately clear where to focus during the second and third rounds of revision.

What Revision Rounds Actually Look Like

The first draft is usually a sprint. I translate quickly straight through so I don't lose the flow of the original. At this stage, sentences with fidelity scores of 7-8 but readability scores of 3-4 are everywhere, because I've carried over the Marathi sentence structure almost as is.

In the second round, I pull out only the sentences with readability scores of 5 or below and rewrite them. The most common task here is splitting sentences. Marathi novels, especially older works, often have a single sentence stretching across four or five lines. Translate that directly into English and the reader can't breathe. But if you chop everything short without thinking, the rhythm of the original disappears. This tug-of-war is the essence of the second round.

The third round is the most demanding. I look at fidelity and readability scores simultaneously and zero in on sentences where both are low. At this stage, I read the source text aloud, and I read my translation aloud too. The walls of my Chennai apartment are thin, so the neighbours knock on the wall sometimes, but it's a step I can't give up. You can only feel whether a sentence carries the right weight when you hear it spoken.

A fourth round (or more) is sometimes necessary. These are usually sentences tangled in cultural context. Say there's a scene where a character reveals emotion while mentioning a food eaten during a specific festival. Do you keep the food's name as is and add a footnote, or do you replace it with an analogous context an English reader would understand? Choices like these can't be resolved by spreadsheet scores. In the end, they get decided through email exchanges with the publisher's editor.

How to Negotiate with a Deadline

Honestly, I'd love to bring every sentence to a fidelity of 9 and a readability of 9. Reality doesn't work that way. There are deadlines, there are physical limits, and if I get stuck on one chapter, the next one falls behind.

Here's the benchmark I use: if the sum of fidelity and readability is 12 or above, I let it pass. Combinations like 7+6, 6+7, 8+5. Any sentence with a combined score of 11 or below gets revisited without exception. If the combined score is 14 or above, I don't touch that sentence. It's already good enough.

Is this benchmark perfect? Of course not. But to finish a chapter with over 300 sentences before the deadline, you need a way to decide quickly where to spend your time. The spreadsheet is exactly that.

Sometimes, after I send the final file to the editor, I receive the proofs about a week later and spot an awkward expression I missed. That stings. But when I open the spreadsheet again, those sentences almost always scored around 11 or 12, right on the borderline. I sometimes consider raising the cutoff to 13 for the next project, but that would add more revision time. This cycle will probably continue for as long as I do this work.

Right now, on my desk, there's a printout of the source text for the next chapter, and a blank spreadsheet template is open. The coffee cup is empty. Time to go refill it.