One verse, every step
Acts 10:9, traced through the whole pipeline — the people who touch it and the shape the text takes at each stage.
Workflows and How does it work? describe this process in the abstract. Here's what it actually looks like on a real verse, start to finish.
- 1
Source
The starting point is the source text alongside translator's notes drawn from established Scripture resources — spelling out what the source leaves implicit.
About noon the next day those [three men] were traveling [along the road] and were coming near the city [of Joppa].
- 2
Draft
A drafter writes a first plain-English semantic paraphrase, bracketing the implicit information the source text leaves unsaid.
On the next day at about noon [when those people already were near Joppa] Peter go-up to the roof of the house [that Peter was staying at] [in-order-to] pray.
- 3
Review
A reviewer and a checker — with Old and New Testament specialists weighing in — refine the draft together until they agree it faithfully carries the meaning, alongside a plain-language backtranslation for a final sanity check.
- 4
Encode
The signed-off draft is parsed into the system's formal semantic representation, flagged for anything ambiguous or incomplete until it checks out clean.
[NP Peter(N)] [VP see-D(V)] [NP dream/vision(N)] . ] | [C [NP at-B(Adp) [AdjP about(Adj)] noon(N)] [NP Peter(N)] [VP go-up(V)] [NP roof(N) [NP -Part Whole(Adp) house(N)]] [NP on-E(Adp) [AdjP next(Adj)] day(N)] [C when(Adp) [NP person(N)] [VP be-F(V)]] [NP near(Adp) Joppa(N)] [AdvP already(Adv)]] [C in-order-to(Adp) [NP Peter(N)] [VP pray(V)]] . ]
- 5
Generate
That same representation — shared across every target language — drives a target language's own lexicon and grammar rules to produce a full draft, ready for a native speaker to take from there.
This walks through the conventional path start to finish. See Workflows for how an AI assistant can shorten the drafting step.