University Homepage School Homepage School Contacts School Search

Lecture 15 - CTC and end-to-end systems

End-to-end systems are systems which learn to directly map from an input sequence X to an output sequence Y , estimating P(Y|X). Sequence discriminative training of HMMs is one way to train such a system. Alternative approaches can be based on recurrent networks which directly map input to output sequences - CTC (Connectionist Temporal Classification) and encoder-decoder architectures. In this lecture we looked at CTC-based systems.

• Baidu's Deep Speech (now also implemented by Mozilla) is an example of an end-to-end system using CTC which maps from acoustic features to a character sequence, with close to state-of-the-art results.
• CTC may be used to train an RNN to map an input sequence to an output sequence (typically of a different length) without requiring frame-level alignment (matching each input frame to an output token)
• CTC is "alignment-free" by summing over all possible alignments.
• Gradient descent training can be carried out as it is possible to back-propagate gradients through CTC
• CTC introduces an additional blank output symbol, which allows an input frame to be not mapped to an output symbol (useful for silence, noise, ...)
• CTC loss function sums over possible alignments
• CTC has a conditional independence assumption: Given the inputs, each output is independent of the other outputs.
• CTC can be interpreted as an HMM with additional (skippable) blank states, trained discriminatively

---

Copyright (c) University of Edinburgh 2015-2018
The ASR course material is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License
licence.txt
This page maintained by Steve Renals.
Last updated: 2018/04/30 21:27:31UTC