MT Post-Edit Quality Reviewer

The MT Post-Edit Quality Reviewer is a specialized AI assistant for assessing the output of machine translation post-editing (MTPE) workflows. As MT becomes a dominant part of the translation supply chain, ensuring that post-editors have done their job correctly — neither over-editing high-quality MT output nor leaving critical errors uncorrected — has become a core quality assurance task. This assistant automates and standardizes that review.

The assistant compares three layers simultaneously: the original source text, the raw MT output, and the post-edited final version. From this triangular comparison, it detects over-editing (where the post-editor changed correct MT output unnecessarily, wasting time and potentially introducing new errors), under-editing (where significant MT errors were left uncorrected and passed into the final deliverable), and net-new errors (mistakes introduced by the post-editor that were not present in either the source or raw MT).

This makes the tool invaluable for language service providers managing MTPE productivity, for MT engine trainers who need clean data to identify where engines fail versus where post-editors fail, and for project managers who need objective evidence to settle disputes about work quality or invoicing.

The assistant produces a segment-level report that highlights each type of finding with a severity tag, a clear explanation, and the three-way text comparison. It also generates aggregate statistics: over-edit rate, under-edit rate, and post-editor error introduction rate. These metrics plug directly into MTPE productivity and quality dashboards.

Ideal use cases include periodic auditor sampling of post-editor portfolios, onboarding assessments for new MTPE vendors, and post-project quality reviews. The assistant works across language pairs and content types, though domain-specific terminology review benefits from a provided glossary.