Multi-Reference Training with Pseudo-References for Neural Translation and Text Generation
This addresses the issue of underutilizing multiple references during training for neural text generation tasks like machine translation and image captioning, offering a novel approach to enhance model performance.
The paper tackles the problem of limited reference diversity in neural text generation training by proposing a method to generate exponentially many pseudo-references from existing human references, resulting in improvements of +1.5 BLEU in machine translation and +3.1 BLEU / +11.7 CIDEr in image captioning.
Neural text generation, including neural machine translation, image captioning, and summarization, has been quite successful recently. However, during training time, typically only one reference is considered for each example, even though there are often multiple references available, e.g., 4 references in NIST MT evaluations, and 5 references in image captioning data. We first investigate several different ways of utilizing multiple human references during training. But more importantly, we then propose an algorithm to generate exponentially many pseudo-references by first compressing existing human references into lattices and then traversing them to generate new pseudo-references. These approaches lead to substantial improvements over strong baselines in both machine translation (+1.5 BLEU) and image captioning (+3.1 BLEU / +11.7 CIDEr).