Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis

Abstract: 

Despite the ability to produce human-level speech for in-domain text, attention-based end-to-end text-to-speech (TTS) systems suffer from text alignment failures that increase in frequency for out-of-domain text. We show that these failures can be addressed using simple location-relative attention mechanisms that do away with content-based query/key comparisons. We compare two families of attention mechanisms: location-relative GMM-based mechanisms and additive energy-based mechanisms. We suggest simple modifications to GMM-based attention that allow it to align quickly and consistently during training, and introduce a new location-relative attention mechanism to the additive energy-based family, called Dynamic Convolution Attention (DCA). We compare the various mechanisms in terms of alignment speed and consistency during training, naturalness, and ability to generalize to long utterances, and conclude that GMM attention and DCA can generalize to very long utterances, while preserving naturalness for shorter, in-domain utterances.

up
0 users have voted:

Paper Details

Authors:
Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby
Submitted On:
14 May 2020 - 6:30pm
Short Link:
Type:
Presentation Slides
Event:
Presenter's Name:
Eric Battenberg
Paper Code:
SPE-L5.4
Document Year:
2020
Cite

Document Files

Location-Relative Attention (slides).pdf

(46)

Subscribe

[1] Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby, "Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis", IEEE SigPort, 2020. [Online]. Available: http://sigport.org/5324. Accessed: Sep. 26, 2020.
@article{5324-20,
url = {http://sigport.org/5324},
author = {Eric Battenberg; RJ Skerry-Ryan; Soroosh Mariooryad; Daisy Stanton; David Kao; Matt Shannon; Tom Bagby },
publisher = {IEEE SigPort},
title = {Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis},
year = {2020} }
TY - EJOUR
T1 - Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis
AU - Eric Battenberg; RJ Skerry-Ryan; Soroosh Mariooryad; Daisy Stanton; David Kao; Matt Shannon; Tom Bagby
PY - 2020
PB - IEEE SigPort
UR - http://sigport.org/5324
ER -
Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby. (2020). Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis. IEEE SigPort. http://sigport.org/5324
Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby, 2020. Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis. Available at: http://sigport.org/5324.
Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby. (2020). "Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis." Web.
1. Eric Battenberg, RJ Skerry-Ryan, Soroosh Mariooryad, Daisy Stanton, David Kao, Matt Shannon, Tom Bagby. Location-Relative Attention Mechanisms For Robust Long-Form Speech Synthesis [Internet]. IEEE SigPort; 2020. Available from : http://sigport.org/5324