SDI’s Language Automation Portal

A new solution for leveraging linguistic automation services

By Alexis Martinez, VP Research & Innovation

With the growth in SDI Media’s latest Machine Translation implementation came the need for an automation pipeline to process large volumes of content for localization with challenging, yet common requirements:

•  Fast, easy and transparent

•  Available to all of SDI’s global facilities

•  Integrated with multiple internal and external services

•  Scalable

•  Developed by a small team

SDI’s Research & Innovation group took this challenge as an opportunity to build a new serverless platform, the Language Automation Portal.

From a user interface perspective, it had to be fast, simple to operate, and provide our global users transparency into the process. It also needed flexibility in handling different inputs and services, and scalability to handle large numbers of Machine Translation and other linguistic automation processes at scale.

Built on AWS serverless technologies, the Language Automation Portal (LAP) is fast, resilient, scalable, and easy to use.  It offers a variety of services including Machine Translation and Chinese conversions, and soon will incorporate SDI’s Auto-Timing for subtitling.

Machine Translation

LAP handles Machine Translation for dubbing, subtitling and metadata.

Dub Scripts and Subtitles are parsed, dialog is Machine Translated, and then merged back to a merged source-target output for use by Subtitle Post Editors, Dubbing Adaptors and Directors.

Language-specific post-processing rules reduce the post-editing effort by handling unit conversions, and other common protocols.

Chinese Conversions

LAP quickly converts Chinese Simplified to Chinese Traditional, and vice versa for subtitling and other services.

Auto-Timing

SDI’s proprietary auto-timing for subtitling aligns dialog to media and applies complex rules for subtitling to produce a clean draft of an English Master Template which is then post-edited and used for localization. Soon this service will be directly integrated into LAP allowing for a seamless pipeline from English Master Template creation to Machine Translation through GTS Pro.

Integrations

The LAP’s backend is integrated with multiple internal and external services and enables SDI’s production management systems to drive SDI’s AI services through the production pipeline from end-to-end:

Technology

LAP is a serverless platform built on multiple AWS technologies including:

  • •  AWS Lambda – serverless compute service
  • •  AWS Step Functions – serverless orchestration service
  • •  AWS Amplify – framework for web/mobile apps, manages other AWS services
    • •  S3 – for front-end hosting and any file storage needed by the app
    • •  DynamoDB – a NoSQL database
    • •  AppSync – a GraphQL API service that sits in front of DynamoDB
    • •  Elasticsearch – for fast searches through our database
    • •  Cognito – for user authentication

The dev effort was substantially reduced by utilizing AWS Amplify and Step Functions. Amplify provided a quick and easy way to spin up and configure other AWS services for our serverless front- and back-end such as S3, DynamoDB and AppSync.  LAP’s Step Function architecture serves as a robust workflow engine – a key component in a solution that’s fast, scalable and reliable.

Key Business Benefits

  • •  Provides Ops with easy access to Language Automation services
  • •  Fast, resilient, scalable
  • •  Integrated with production management systems for Machine Assisted Translation automation

 

Roadmap

LAP’s service offerings will continue to expand throughout 2021 as SDI’s AI services continue to grow including more Machine Translation languages, speech-to-text, glossary handling, and additional automation technology integrations.