A Simple Baseline for Beam Search Reranking
This work provides an incremental improvement for machine translation researchers by offering a decoupled baseline to evaluate reranking methods.
The paper tackles the gap between evaluation metrics and learning algorithms in machine translation by proposing a simple reranking method that predicts BLEU scores without extra data or parameters, serving as a clean baseline for future research.
Reranking methods in machine translation aim to close the gap between common evaluation metrics (e.g. BLEU) and maximum likelihood learning and decoding algorithms. Prior works address this challenge by training models to rerank beam search candidates according to their predicted BLEU scores, building upon large models pretrained on massive monolingual corpora -- a privilege that was never made available to the baseline translation model. In this work, we examine a simple approach for training rerankers to predict translation candidates' BLEU scores without introducing additional data or parameters. Our approach can be used as a clean baseline, decoupled from external factors, for future research in this area.