Sequence-to-Sequence Learning as Beam-Search Optimization
This work addresses biases in seq2seq training for NLP practitioners, offering a structured approach that unifies training and test-time usage, though it is incremental as it builds on existing seq2seq architectures.
The paper tackled the problem of local training biases in sequence-to-sequence models by introducing a beam-search training scheme that learns global sequence scores, resulting in outperforming a highly-optimized attention-based seq2seq system on tasks like word ordering, parsing, and machine translation.
Sequence-to-Sequence (seq2seq) modeling has rapidly become an important general-purpose NLP tool that has proven effective for many text-generation and sequence-labeling tasks. Seq2seq builds on deep neural language modeling and inherits its remarkable accuracy in estimating local, next-word distributions. In this work, we introduce a model and beam-search training scheme, based on the work of Daume III and Marcu (2005), that extends seq2seq to learn global sequence scores. This structured approach avoids classical biases associated with local training and unifies the training loss with the test-time usage, while preserving the proven model architecture of seq2seq and its efficient training approach. We show that our system outperforms a highly-optimized attention-based seq2seq system and other baselines on three different sequence to sequence tasks: word ordering, parsing, and machine translation.