Semantic Label Smoothing for Sequence to Sequence Problems
This addresses regularization and overfitting in seq2seq models, particularly for machine translation, though it appears incremental relative to existing token-level or random substitution approaches.
The paper tackles the challenge of applying label smoothing to sequence-to-sequence problems like machine translation, where the large output space makes traditional approaches intractable. The proposed method smooths over semantically similar and well-formed sequences, achieving consistent and significant improvements over state-of-the-art techniques on various datasets.
Label smoothing has been shown to be an effective regularization strategy in classification, that prevents overfitting and helps in label de-noising. However, extending such methods directly to seq2seq settings, such as Machine Translation, is challenging: the large target output space of such problems makes it intractable to apply label smoothing over all possible outputs. Most existing approaches for seq2seq settings either do token level smoothing, or smooth over sequences generated by randomly substituting tokens in the target sequence. Unlike these works, in this paper, we propose a technique that smooths over \emph{well formed} relevant sequences that not only have sufficient n-gram overlap with the target sequence, but are also \emph{semantically similar}. Our method shows a consistent and significant improvement over the state-of-the-art techniques on different datasets.