Multitask Learning with CTC and Segmental CRF for Speech Recognition

Multitask Learning with CTC and Segmental CRF for Speech Recognition

Liang Lu, Lingpeng Kong, Chris Dyer, Noah A. Smith

Intro

segmental conditional random fields（SCRFs）和connectionist temporal classification（CTC）是两个给序列end-to-end labeling方法。这篇paper就是将两者结合起来进行优化，共享RNN encoder。

TBC