Sixteen Suggested Edits, Two Accepted

A while back I proofread a colleague's English translation of a marital-status certificate from a civil registry office in North Sumatra. Before reading it myself, I ran an SI-assisted first pass, which came back with sixteen suggested edits.

I kept two.

What I accepted

The first was a capital letter: "a certificate" became "a Certificate", where the word names the document being requested. Small, but in an official text the capital tells the reader which certificate we mean.

The second was Kelurahan. The translator had written "Urban Village". I changed it to "Sub-District". That one is not taste. In the chain of administrative units, Sub-District sits where the reader expects it.

I also added one edit of my own. In the closing certification sentence, "up to the present" became "to date". The same Indonesian phrase appears earlier in the grounds, and there "up to the present" reads fine. In a closing formula, "to date" is the more usual English.

What I rejected

The other fourteen were not errors. The address was written in the translator's own format, with no "Indonesia" added. The signatory's titles appeared without a bracketed gloss. "Statement" was suggested to become "Declaration Letter". The rank of the official was to be reworded, and so was the name of the e-signature authority in the footer.

Each of these is a defensible choice, and some match my own house style. None of them was wrong. So they stayed.

One suggestion would have made the text worse. The source said the person belum pernah kawin sejak bercerai, and the translator wrote "has not remarried since his divorce". The pass proposed "married". That loses the meaning. "Remarried" is exactly right.

Another flagged "under religious or civil law or customary practices" as needing work. It did not. It was already correct.

The rule I take from it

Proofreading is the correction of errors. If I cannot say what is wrong with a phrase, only that I would have written it differently, I leave it.

The tools make this harder, oddly. An automated pass will happily suggest sixteen changes, and each one looks reasonable on its own. The proofreader's job is to say no to most of them. Two out of sixteen is roughly what a careful review looks like.