Machine Learning

Implementing Multilingual Support with Amazon Translate - Real-Time Translation and Custom Terminology

Learn about real-time translation with Translate, improving translation quality with custom terminology, and leveraging batch translation for large-scale document processing.

About 7 min read

Translate Overview

Translate is a service that provides neural machine translation across 75+ languages, processing up to 100 KB of text per request. The real-time translation API (TranslateText) is a synchronous call that returns results instantly, making it suitable for integration with chat and web forms. The response includes automatic source language detection in addition to translated text, allowing it to handle cases where the input language is unknown. Batch translation (StartTextTranslationJob) executes bulk translation of large document volumes as asynchronous jobs, using S3 as storage. Two translation modes are available - Standard and Active Custom Translation - allowing you to balance accuracy and cost based on your use case.

Custom Terminology and Batch Translation

Custom terminology defines term pairs in CSV or TMX format, ensuring specified terms are always used during translation. This lets you control translations such as preventing the product name "Amazon Aurora" from being translated and keeping it in its original form. You can create up to 5,000 terminology resources per account, with a maximum size of 10 MB per resource. DirectionalTerminology lets you define different mappings per translation direction, enabling fine-grained control such as applying different translations for Japanese-to-English versus English-to-Japanese. Batch translation places documents in an S3 bucket input folder and stores translation results in an output folder. It supports HTML, DOCX, XLIFF, PowerPoint, and Excel formats, preserving tag structure during translation. Each job can process up to 5 million characters, and the number of concurrent jobs is limited by per-region default quotas.

Parallel Data and Active Custom Translation

Parallel data (parallel corpora) provides source and translation text pairs in CSV or TMX format, adapting the translation model to domain-specific expressions. Translation quality improves significantly in fields with heavy specialized terminology, such as technical documentation, legal documents, and medical documents. Active Custom Translation (ACT) uses parallel data to adjust the model in real time, generating more natural, context-aware translations than custom terminology alone. Parallel data supports a minimum of 10 sentence pairs, but the effect becomes pronounced with several thousand or more pairs. Profanity masking automatically detects inappropriate expressions and masks or excludes them. Formality settings let you control the formality level (polite/casual) of translations, allowing you to use appropriate tones for business documents versus casual chat. Supported languages for Formality are limited to Japanese, German, French, Italian, Portuguese, and others, so verify support for your target language pair in advance. For a systematic understanding of Translate from basics to advanced usage, books on Amazon are a great resource.

Use Cases and Architecture Patterns

Translate applies to a wide range of workloads. For multilingual e-commerce product descriptions, a serverless pipeline using EventBridge + Lambda + Translate that auto-translates on new product registration and stores results in DynamoDB is effective. In customer support, combining it with Amazon Connect enables real-time chat message translation so agents and users can converse in different languages. For knowledge base localization, batch translation processes technical documents stored in S3, and Amazon Kendra provides multilingual search. In social media monitoring, a real-time pipeline receives posts via Kinesis Data Streams, unifies text to English with Translate, then runs sentiment analysis with Comprehend.

Comparison with Other Translation Services

Translate's strength lies in its integration with the AWS ecosystem. It supports fine-grained access control via IAM, API auditing via CloudTrail, and private connectivity via VPC endpoints, meeting enterprise security requirements. Compared to Google Cloud Translation, ACT enables parallel data adaptation at an additional training cost - Google's Glossary is equivalent to custom terminology but lacks a context-adaptive model customization feature. DeepL API offers high translation quality for a limited set of language pairs, but Translate surpasses it in coverage with 75+ languages and broader Formality setting support. Compared to on-premises neural translation, Translate eliminates GPU infrastructure management and automates scaling. However, weakness in literary expressions and idioms is a common challenge across all cloud translation services.

Translate Pricing

Translate pricing is based on the number of characters translated, with real-time translation costing approximately $15 per million characters. Batch translation uses the same per-character rate but allows bulk processing of large document volumes via S3. The first 12 months include a free tier of 2 million characters per month. There is no additional charge for using custom terminology. Active Custom Translation incurs additional charges for parallel data training, billed by training time. You can optimize costs by pre-stripping HTML tags and metadata from translation targets to reduce non-translatable character counts. Adding a caching layer to avoid re-translating identical text is also effective. Each request has a minimum charge of 15 characters, so designs that send extremely short strings in high volume are inefficient.

Summary

Translate is a service that enables multilingual support across 75+ languages using neural machine translation. Custom terminology controls the translation of product names and specialized terms, while Active Custom Translation provides domain adaptation through parallel data. Formality settings let you switch between polite and casual tones, and batch translation automates the multilingual conversion of large content volumes. Integration with the AWS ecosystem supports a wide range of use cases from serverless auto-translation pipelines to real-time chat translation.

Related Services

Amazon TranslateA neural machine translation service for multilingual text translationAmazon S3A scalable object storage serviceAWS LambdaA serverless compute service that runs code without server management

Related Articles

Implementing Text-to-Speech with Amazon Polly - Neural Voices and SSML Speech ControlGenerate natural-sounding speech with the Neural TTS engine and control speech rate, pitch, and pauses with SSML tags. Learn how to build diverse audio content using real-time streaming and asynchronous synthesis to S3.Practicing Infrastructure as Code with AWS CloudFormation - Template Design and Stack ManagementDefine infrastructure with templates and preview impact with change sets before deployment. Detect configuration drift with drift detection and deploy across your entire organization with StackSets.Building a Real-Time Data Pipeline with Amazon Kinesis - Choosing Between Data Streams and Data FirehoseIngest real-time data with Data Streams and automatically deliver it to S3, Redshift, and OpenSearch with Data Firehose. This article explains how to build a streaming pipeline with shard design and on-demand mode selection.

More on This Topic

GPU-Based Machine Learning Training with AWS Batch - Cost-Efficient Large-Scale TrainingRun GPU training with your existing Docker containers, and cut costs by up to 90% using Spot Instances and checkpointing. Includes guidance on when to choose Batch over SageMaker.Using Claude on Amazon Bedrock - Model Selection, Prompt Design, and Cost OptimizationCompares the Anthropic Claude models available on Amazon Bedrock, provides model selection guidelines by use case, and covers prompt design best practices and cost optimization.Building RAG Applications with Amazon Bedrock Knowledge Bases - Implementing Retrieval-Augmented GenerationAutomatically index documents on S3 and unify search and generation with the RetrieveAndGenerate API. Covers chunking strategy selection and safety enforcement with Guardrails.Getting Started with Quantum Computing on Amazon Braket - Designing and Simulating Quantum CircuitsPrototype for free with local simulators, then run quantum circuits on IonQ and Rigetti hardware. Covers implementing VQE and QAOA with hybrid jobs.Privacy-Preserving ML with AWS Clean Rooms ML - Build Models Without Sharing DataLearn how to build lookalike models with Clean Rooms ML, apply differential privacy, and leverage the results for ad targeting.Implementing Natural Language Processing with Amazon Comprehend - Sentiment Analysis and Entity ExtractionLearn about sentiment analysis, entity extraction, and building custom classification models with Comprehend.Building Conversational Bots - Natural Conversation Interfaces with Amazon Lex and PollyLearn how to build conversational bots using Amazon Lex and Amazon Polly.Document Text Extraction - Intelligent Document Processing with Amazon TextractLearn how to automatically extract text, tables, and form data from documents with Amazon Textract, and build natural language processing pipelines by integrating with Amazon Comprehend. This article covers automation patterns for invoice processing and contract analysis.

Similar Articles and Services

Amazon TranslateA neural machine translation service supporting over 75 languagesA global company wants to automatically translate product manuals written in English into Japanese, French, and German. They need to translate a large volume of documents in batch. Which AWS service is most appropriate?An insurance company is undertaking a project to digitize a large volume of invoices and application forms. They want to automatically extract tabular data and form field values from scanned PDFs. Which AWS service is most appropriate?Fundamentals of AI Solutions (AWS Certified AI Practitioner)In this domain, you learn the main AI/ML services that AWS provides. The central theme is knowing when to use Amazon Bedrock (a foundation model API), Amazon SageMaker (a machine learning platform), Amazon Q (a business-oriented assistant), and purpose-specific AI services (Comprehend, Rekognition, Transcribe, and others). It is the largest domain, accounting for 28% of the AIF-C01 exam.