Skip to content

[Paper Note] Experts, Errors, and Context: A Large-Scale Study of Human Evaluation for Machine Translation, Markus Freitag+, arXiv'21, 2021.04 #1221

Description

@AkihikoWatanabe

URL

Authors

  • Markus Freitag
  • George Foster
  • David Grangier
  • Viresh Ratnakar
  • Qijun Tan
  • Wolfgang Macherey

Abstract

  • Human evaluation of modern high-quality machine translation systems is a difficult problem, and there is increasing evidence that inadequate evaluation procedures can lead to erroneous conclusions. While there has been considerable research on human evaluation, the field still lacks a commonly-accepted standard procedure. As a step toward this goal, we propose an evaluation methodology grounded in explicit error analysis, based on the Multidimensional Quality Metrics (MQM) framework. We carry out the largest MQM research study to date, scoring the outputs of top systems from the WMT 2020 shared task in two language pairs using annotations provided by professional translators with access to full document context. We analyze the resulting data extensively, finding among other results a substantially different ranking of evaluated systems from the one established by the WMT crowd workers, exhibiting a clear preference for human over machine output. Surprisingly, we also find that automatic metrics based on pre-trained embeddings can outperform human crowd workers. We make our corpus publicly available for further research.

Translation (by gpt-5-nano)

  • 現代の高品質な機械翻訳システムの人間評価は難題であり、不十分な評価手法が誤った結論につながるというエビデンスが高まっている。人間評価については相当な研究が進んでいるものの、この分野には依然として広く受け入れられた標準的な手順は欠如している。これを目標に一歩進めるものとして、明示的な誤り分析に基づく評価方法論を提案する。これは多次元品質指標(MQM)フレームワークに基づくものである。これまでで最大規模のMQM研究を実施し、WMT 2020 共有タスクの上位システムの出力を、全文の文脈にアクセス可能なプロの翻訳者が提供した注釈を用いて、二言語ペアで評価した。得られたデータを広範に分析し、他の結果とともに、WMTクラウドワーカーが確立した評価システムのランキングとはかなり異なるランキングを示し、人間の出力を機械出力よりも明確に好む傾向があることを示した。驚くべきことに、事前学習済み埋め込みに基づく自動指標が人間のクラウドワーカーよりも優れていることも明らかになった。さらに研究のためにコーパスを公開する。

Summary (by gpt-4o-mini)

  • 機械翻訳の人間評価は難しく、標準的な手法が不足している。そこで、誤り分析に基づく評価方法論を提案し、MQMフレームワークを用いてWMT 2020の上位システム出力をプロの翻訳者による注釈で評価。分析の結果、WMTクラウドワーカーのランキングと異なる結果が得られ、人間が機械出力よりも人間の出力を好む傾向を示した。さらに、自動指標がクラウドワーカーよりも優れたことも判明し、研究用コーパスを公開。

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions