Macro-Average: Rare Types Are Important Too

TL;DR

This paper investigates the use of the MacroF1 metric, a simple type-based classifier, for evaluating machine translation quality, demonstrating its competitiveness and ability to reveal differences in translation methods.

Contribution

The study introduces MacroF1 as an effective, transparent alternative to model-based MT metrics, applicable across domains and languages, and capable of highlighting qualitative differences.

Findings

MacroF1 is competitive on direct assessment.

MacroF1 outperforms other metrics in cross-lingual retrieval tasks.

MacroF1 reveals qualitative differences between translation methods.

Abstract

While traditional corpus-level evaluation metrics for machine translation (MT) correlate well with fluency, they struggle to reflect adequacy. Model-based MT metrics trained on segment-level human judgments have emerged as an attractive replacement due to strong correlation results. These models, however, require potentially expensive re-training for new domains and languages. Furthermore, their decisions are inherently non-transparent and appear to reflect unwelcome biases. We explore the simple type-based classifier metric, MacroF1, and study its applicability to MT evaluation. We find that MacroF1 is competitive on direct assessment, and outperforms others in indicating downstream cross-lingual information retrieval task performance. Further, we show that MacroF1 can be used to effectively compare supervised and unsupervised neural machine translation, and reveal significant qualitative differences in the methods' outputs.