English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Neural machine translation (NMT) models are conventionally trained withtoken-level negative log-likelihood (NLL), which does not guarantee that thegenerated translations will be optimized for a selected sequence-levelevaluation metric. Multiple approaches are proposed to train NMT with BLEU asthe reward, in order to directly improve the metric. However, it was reportedthat the gain in BLEU does not translate to real quality improvement, limitingthe application in industry. Recently, it became clear to the community thatBLEU has a low correlation with human judgment when dealing withstate-of-the-art models. This leads to the emerging of model-based evaluationmetrics. These new metrics are shown to have a much higher human correlation.In this paper, we investigate whether it is beneficial to optimize NMT modelswith the state-of-the-art model-based metric, BLEURT. We propose acontrastive-margin loss for fast and stable reward optimization suitable forlarge NMT models. In experiments, we perform automatic and human evaluations tocompare models trained with smoothed BLEU and BLEURT to the baseline models.Results show that the reward optimization with BLEURT is able to increase themetric scores by a large margin, in contrast to limited gain when training withsmoothed BLEU. The human evaluation shows that models trained with BLEURTimprove adequacy and coverage of translations. Code is available viahttps://github.com/naver-ai/MetricMT.

Reward Optimization for Neural Machine Translation with Learned Metrics