 |
ML Newsnew | past | best | rss | submit | about |
| A Test Suite and Manual Evaluation of Document-Level NMT at WMT19 (arxiv.org) | |
| 1 point by sel1 on Aug 10, 2019 | hide | past | pdf | discuss on HN | |
|
In plain words: A new test set of passages probes cross-sentence issues like pronouns and consistent word choice, so translation systems can be judged on whole documents rather than isolated sentences. Hand-checking the outputs revealed specific error types that only appear at document level. AbstractAs the quality of machine translation rises and neural machine translation (NMT) is moving from sentence to document level translations, it is becoming increasingly difficult to evaluate the output of translation systems. We provide a test suite for WMT19 aimed at assessing discourse phenomena of MT systems participating in the News Translation Task. We have manually checked the outputs and identified types of translation errors that are relevant to document-level translation. | |
| add comment on HN |