Classical Structured Prediction Losses for Sequence to Sequence Learning
This work addresses the problem of improving sequence-level training for neural models in tasks like translation and summarization, but it is incremental as it adapts existing classical methods to modern architectures.
The paper applied classical structured prediction losses to neural sequence-to-sequence models, achieving slightly better performance than beam search optimization and reporting new state-of-the-art results on IWSLT'14 German-English translation and Gigaword summarization, with 41.5 BLEU on WMT'14 English-French translation.
There has been much recent work on training neural attention models at the sequence-level using either reinforcement learning-style methods or by optimizing the beam. In this paper, we survey a range of classical objective functions that have been widely used to train linear models for structured prediction and apply them to neural sequence to sequence models. Our experiments show that these losses can perform surprisingly well by slightly outperforming beam search optimization in a like for like setup. We also report new state of the art results on both IWSLT'14 German-English translation as well as Gigaword abstractive summarization. On the larger WMT'14 English-French translation task, sequence-level training achieves 41.5 BLEU which is on par with the state of the art.