Machine translation (MT) and post-editing (PE) can be applied to many different content types and use cases. Measuring MT output quality through testing is an effective way to predict business outcomes. How do we test in the era of neural machine translation (NMT)? Metrics that were useful for statistical machine translation are much less reliable for NMT, where fluent and context-sensitive output may deviate from gold-standard reference translations. This means that MT testing needs to be re-evaluated in the context of NMT.
Download this white paper to learn best practices for these key MT and PE testing aspects:
- Assessing content
- Creating representative test sets
- Automated and human testing
- Analyzing test results
Please fill in your details below to download the White paper