On the Inference Calibration of Neural Machine Translation

TL;DR

This paper investigates the calibration of neural machine translation models during inference, revealing persistent miscalibration issues and proposing a graduated label smoothing method to enhance both calibration and translation quality.

Contribution

It provides an in-depth analysis of NMT calibration challenges during inference and introduces a novel graduated label smoothing technique to improve calibration and translation performance.

Findings

Miscalibration remains a severe challenge during inference.

Calibration correlates with translation performance and linguistic properties.

Graduated label smoothing improves calibration and translation quality.

Abstract

Confidence calibration, which aims to make model predictions equal to the true correctness measures, is important for neural machine translation (NMT) because it is able to offer useful indicators of translation errors in the generated output. While prior studies have shown that NMT models trained with label smoothing are well-calibrated on the ground-truth training data, we find that miscalibration still remains a severe challenge for NMT during inference due to the discrepancy between training and inference. By carefully designing experiments on three language pairs, our work provides in-depth analyses of the correlation between calibration and translation performance as well as linguistic properties of miscalibration and reports a number of interesting findings that might help humans better analyze, understand and improve NMT models. Based on these observations, we further propose a new graduated label smoothing method that can improve both inference calibration and translation performance.