Confidence-Aware Scheduled Sampling for Neural Machine Translation

TL;DR

This paper introduces confidence-aware scheduled sampling for neural machine translation, which dynamically adjusts training exposure based on model confidence, leading to improved translation quality and faster convergence.

Contribution

It proposes a novel confidence-based schedule strategy for scheduled sampling, enhancing performance and training efficiency in neural machine translation.

Findings

Significantly outperforms vanilla scheduled sampling in translation quality.

Achieves faster convergence during training.

Effective across multiple language pairs and Transformer models.

Abstract

Scheduled sampling is an effective method to alleviate the exposure bias problem of neural machine translation. It simulates the inference scene by randomly replacing ground-truth target input tokens with predicted ones during training. Despite its success, its critical schedule strategies are merely based on training steps, ignoring the real-time model competence, which limits its potential performance and convergence speed. To address this issue, we propose confidence-aware scheduled sampling. Specifically, we quantify real-time model competence by the confidence of model predictions, based on which we design fine-grained schedule strategies. In this way, the model is exactly exposed to predicted tokens for high-confidence positions and still ground-truth tokens for low-confidence positions. Moreover, we observe vanilla scheduled sampling suffers from degenerating into the original teacher forcing mode since most predicted tokens are the same as ground-truth tokens. Therefore, under the above confidence-aware strategy, we further expose more noisy tokens (e.g., wordy and incorrect word order) instead of predicted ones for high-confidence token positions. We evaluate our approach on the Transformer and conduct experiments on large-scale WMT 2014 English-German, WMT 2014 English-French, and WMT 2019 Chinese-English. Results show that our approach significantly outperforms the Transformer and vanilla scheduled sampling on both translation quality and convergence speed.