Macro-Average: Rare Types Are Important Too
This addresses the need for transparent, domain-agnostic MT evaluation metrics, particularly for researchers and practitioners dealing with rare types and downstream tasks, though it is incremental as it adapts an existing metric to a new application.
The paper tackled the problem of evaluating machine translation (MT) by proposing MacroF1, a type-based classifier metric, as an alternative to traditional and model-based metrics. It found that MacroF1 is competitive in direct assessment and outperforms others in cross-lingual information retrieval, while also revealing qualitative differences in supervised vs. unsupervised neural MT outputs.
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.