Align With Purpose: Optimize Desired Properties in CTC Models with a General Plug-and-Play Framework

arXiv:2307.01715v3 Announce Type: replace-cross
Abstract: Connectionist Temporal Classification (CTC) is a widely used criterion for training supervised sequence-to-sequence (seq2seq) models. It enables learning the relations between input and output sequences, termed alignments, by marginalizing over perfect alignments (that yield the ground truth), at the expense of imperfect alignments. This binary differentiation of perfect and imperfect alignments falls short of capturing other essential alignment properties that hold significance in other real-world applications. Here we propose $\textit{Align With Purpose}$, a $\textbf{general Plug-and-Play …

abstract arxiv classification criterion cs.cl cs.lg cs.sd eess.as framework general relations seq2seq temporal training truth type