How to implement open source MT solutions

OUTLINE

The History of Machine Translation

Very brief overview of the history of machine translation including the different approaches.

A Short Introduction to Statistical Machine Translation

A 5-min practical introduction to the principles of SMT (essential for the understanding of the Moses MT process).

Overview of Open Source Machine Translation

Overview of the different initiatives in OSS MT, outlining pros/cons and language pairs already available. A brief current thinking on OSS benefits in a corporate environment.

Data Sources and Data Cleaning – Making the Most of your Translation Memories

• Data Sources
  • Open Data
  • TDA
  • BigTM.net
  • Your Data (Translation Memories)
• Data Formats (TMX etc.)
• Character Encoding
• Parallel Data
  • Data Selection
  • Sentence Splitting and Alignment
    • Language-specific Requirements
• Monolingual Data
  • Data Selection
  • Sentence Splitting
    • Language-specific Requirements
• Data Cleaning
• Special considerations for certain language pairs (morphologically rich languages, East Asian languages)
• Translation Pre- and Post-Processing
  • Dates, Times, Product Names
  • Formatting, upper/lowercasing lost
• Defining the expected quality: Tuning Data and Evaluation Data

The Moses Statistical Machine Translation System and Associated Components

• Moses and required components (incl. licensing considerations)
• Deploying Moses
  • Basic
  • Moses on EC2 (Security Considerations)
  • Performance Considerations (Clustering)

Training, Tuning and Translating with Moses

• High Level Overview of Phases of Moses MT System Training described earlier.
• Tuning
  • Domain adaptation
  • Project adaptation
  • Iterative Improvement of MT Output Quality

Integrating Moses into a Localization Workflow

• Integration Options
• Offline Integration
  • Fuzzy Matching with TM
  • Exporting Untranslated Segments to MT
  • Translating with Moses
  • Importing Translations back into TM system
  • Post-editing MT
    • Different quality levels of post-editing
    • Working with post-editors
    • Typical mistakes of an SMT system
  • Considerations for Integration with Crowdsourcing
• Online Integration
  • Server API
  • Performance Considerations

Getting Support and Supporting the Community

• Support Concepts in OSS
• Contributing to OSS Projects
• Influencing the Direction of OSS Projects