Back office ops · Production

TransPerfect improves translation quality and efficiency using Amazon Bedrock for automated post-editing and transcreation

The problem

AWS and TransPerfect's localization workflow steps—including machine translation post-editing and transcreation—were often manual, costly, and time-consuming, with transcreation especially resistant to automation and dependent on highly skilled human linguists.

Workflow diagram · grounded in source
1
Translation memory lookup
integration
“The translation memory is a client-specific repository of previously translated and approved content. It's always applied first and maximizes the reuse of existing translations.”
2
Machine translation via Amazon Translate
ai_action
“After existing translations are applied, new content is processed through machine translation using Amazon Translate.”
3
LLM automated post-edit
ai_action
“By using style guides, relevant examples of approved translations, and examples of errors to avoid, the LLM is prompted to improve existing machine translations.”
4
Hallucination detection via Guardrails
validation
“Amazon Bedrock supports contextual grounding checks to detect and filter hallucinations if the responses are factually incorrect or inconsistent.”
5
Route to HPE or no-touch workflow
routing
“This post-edited content is either handed off to a linguist for a lighter post-edit (a less difficult task) or is applied in "no human touch workflows" to greatly improve the output.”
6
Human post-edit (lighter touch)
human_review
“A subject matter expert linguist revises and perfects the machine-translated content.”
7
LLM transcreation candidate generation
ai_action
“translations are produced through Anthropic's Claude or Amazon Nova Pro on Amazon Bedrock, with the prompt to create multiple candidate translations with some variations”
8
Human linguist selects best candidate
human_review
“the human linguist chooses the most suitable adapted translation instead of composing it from scratch”
Reported outcome

LLM-powered post-editing and transcreation delivered up to 50% cost savings for translations, up to 60% linguist productivity gains for transcreation, up to 40% cost savings across translation workflows, and up to 80% reduction in project turnaround times.

Reported metrics
LLM edits showing improved translation qualityover 95%
Overall cost savings for translations (post-editing)up to 50%
Linguist productivity gains for transcreationup to 60%
Cost savings within translation workflowsup to 40%
Show all 7 reported metrics
LLM edits showing improved translation qualityover 95%
overall cost savings for translations (post-editing)up to 50%
linguist productivity gains for transcreationup to 60%
cost savings within translation workflowsup to 40%
reduction in project turnaround timesup to 80%
human linguist focus shiftfreeing human linguists for higher-level tasks
translation quality improvementenhanced quality across the board
Reported stack
Amazon BedrockGlobalLinkAmazon TranslateAmazon Bedrock GuardrailsAnthropic's ClaudeAmazon Nova Pro
Source
https://aws.amazon.com/blogs/machine-learning/how-transperfect-improved-translation-quality-and-efficiency-using-amazon-bedrock?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

LLM-powered post-editing and transcreation delivered up to 50% cost savings for translations, up to 60% linguist productivity gains for transcreation, up to 40% cost savings across translation workflows, and up to 80%…

What tools did this team use?

Amazon Bedrock, GlobalLink, Amazon Translate, Amazon Bedrock Guardrails, Anthropic's Claude, Amazon Nova Pro.

What results were reported?

LLM edits showing improved translation quality: over 95%; Overall cost savings for translations (post-editing): up to 50%; Linguist productivity gains for transcreation: up to 60%; Cost savings within translation workflows: up to 40% (source-reported, not independently verified).

How is this back office ops AI workflow structured?

Translation memory lookup → Machine translation via Amazon Translate → LLM automated post-edit → Hallucination detection via Guardrails → Route to HPE or no-touch workflow → Human post-edit (lighter touch) → LLM transcreation candidate generation → Human linguist selects best candidate.

Related back office ops cases