Model & data planning
How much training data do you need for machine translation?
Updated 2026-08-01 · 4 min read
Short answer
For machine translation, volume matters less than composition. Training and evaluating translation between English and Indian languages, and between Indian languages. The corpus profile that works is sentence-aligned parallel corpora, register-matched to your product, enforced terminology glossary. Start with a pilot sized to move human adequacy and fluency scores, terminology compliance rate measurably, confirm the gain on held-out data recorded under deployment conditions, then scale the configuration that worked rather than scaling everything.
Key takeaways
- Success is measured on human adequacy and fluency scores, terminology compliance rate, back-translation divergence.
- The most common failure is pivoting everything through english
- Data profile: sentence-aligned parallel corpora, register-matched to your product, enforced terminology glossary.
What the model actually needs
Training and evaluating translation between English and Indian languages, and between Indian languages.
- Sentence-aligned parallel corpora
- Register-matched to your product
- Enforced terminology glossary
Metrics that tell you when you have enough
Collect against a metric, not against a number of hours. When a pilot batch moves the metric and a second batch of the same profile moves it less, you are at the point where composition, not volume, is the constraint.
- Human adequacy and fluency scores
- Terminology compliance rate
- Back-translation divergence
Common mistakes
- Pivoting everything through English
- Post-edited machine output passed off as human translation
A sensible collection sequence
| Phase | Volume | Purpose |
|---|---|---|
| Pilot | 10–20 hours | Validate format, acoustics and annotation against your pipeline |
| First production batch | 100–300 hours | Move the primary metric and expose composition gaps |
| Targeted top-up | 50–150 hours | Fill the specific dialects, conditions or edge cases the eval exposed |
| Evaluation set | 5–20 hours | Held-out, deployment-condition data never used for training |
Services that supply this data
This use case is normally served by translation and localisation, human data collection for llm. Most programmes combine two of them, because raw collection without matched annotation rarely moves an applied metric on its own.
Hold back an honest evaluation set
Reserve deployment-condition data that never enters training. Teams that evaluate on data recorded in the same sessions as their training data consistently overestimate real-world performance, then discover the gap after launch.
Frequently asked questions
What data profile suits machine translation?
Sentence-aligned parallel corpora, Register-matched to your product, Enforced terminology glossary
Which metrics should we track?
Human adequacy and fluency scores, Terminology compliance rate, Back-translation divergence
What goes wrong most often?
Pivoting everything through English
Can we start small?
Yes. A 10–20 hour pilot delivered in your ingest format is the standard first step, and it usually exposes format or annotation mismatches that would have been expensive at volume.
Related reading
Turn this into a dataset specification
Tell us the languages, speaker count and minutes. You get a written scope, a protocol and a fixed price within one working day.