Star Temporal Classification: Sequence Classification with Partially Labeled Data

TL;DR

This paper introduces Star Temporal Classification (STC), a novel algorithm for sequence classification that effectively handles partially labeled and unsegmented data, significantly improving performance in speech and handwriting recognition tasks.

Contribution

The paper presents STC, a new loss function using a star token for better alignment in partially labeled sequence data, implemented with WFSTs and GTN for automatic differentiation.

Findings

STC recovers most of the supervised performance with up to 70% missing labels.

STC performs well in both speech and handwriting recognition tasks.

The method is adaptable to various sequence classification problems.

Abstract

We develop an algorithm which can learn from partially labeled and unsegmented sequential data. Most sequential loss functions, such as Connectionist Temporal Classification (CTC), break down when many labels are missing. We address this problem with Star Temporal Classification (STC) which uses a special star token to allow alignments which include all possible tokens whenever a token could be missing. We express STC as the composition of weighted finite-state transducers (WFSTs) and use GTN (a framework for automatic differentiation with WFSTs) to compute gradients. We perform extensive experiments on automatic speech recognition. These experiments show that STC can recover most of the performance of supervised baseline when up to 70% of the labels are missing. We also perform experiments in handwriting recognition to show that our method easily applies to other sequence classification tasks.