Word emphasis prediction for expressive text to speech

Yosi Mass; Slava Shechtman; Moran Mordechay; Ron Hoory; Oren Sar Shalom; Guy Lev; David Konopnicki

doi:10.21437/interspeech.2018-1159

Word emphasis prediction for expressive text to speech

Yosi Mass, Slava Shechtman, Moran Mordechay, Ron Hoory, Oren Sar Shalom, Guy Lev, David Konopnicki

IBM

Research output: Contribution to journal › Conference article › peer-review

15 Scopus citations

Abstract

Word emphasis prediction is an important part of expressive prosody generation in modern Text-To-Speech (TTS) systems. We present a method for predicting emphasized words for expressive TTS, based on a Deep Neural Network (DNN). We show that the presented method outperforms machine learning methods based on hand-crafted features in terms of objective metrics such as precision and recall. Using a listening test, we further demonstrate that the contribution of the predicted emphasized words to the expressiveness of the synthesized speech is subjectively perceivable.

Original language	English
Pages (from-to)	2868-2872
Number of pages	5
Journal	Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
Volume	2018-September
DOIs	https://doi.org/10.21437/interspeech.2018-1159
State	Published - 2018
Externally published	Yes
Event	19th Annual Conference of the International Speech Communication, INTERSPEECH 2018 - Hyderabad, India Duration: 2 Sep 2018 → 6 Sep 2018

Bibliographical note

Publisher Copyright:
© 2018 International Speech Communication Association. All rights reserved.

Keywords

Deep learning
Expressive text to speech
Prosody
Speech synthesis
Word emphasis

Access to Document

10.21437/interspeech.2018-1159

Cite this

@article{5fb7794c9c3b4bc7b60d42392a69de7b,

title = "Word emphasis prediction for expressive text to speech",

abstract = "Word emphasis prediction is an important part of expressive prosody generation in modern Text-To-Speech (TTS) systems. We present a method for predicting emphasized words for expressive TTS, based on a Deep Neural Network (DNN). We show that the presented method outperforms machine learning methods based on hand-crafted features in terms of objective metrics such as precision and recall. Using a listening test, we further demonstrate that the contribution of the predicted emphasized words to the expressiveness of the synthesized speech is subjectively perceivable.",

keywords = "Deep learning, Expressive text to speech, Prosody, Speech synthesis, Word emphasis",

author = "Yosi Mass and Slava Shechtman and Moran Mordechay and Ron Hoory and Shalom, {Oren Sar} and Guy Lev and David Konopnicki",

note = "Publisher Copyright: {\textcopyright} 2018 International Speech Communication Association. All rights reserved.; 19th Annual Conference of the International Speech Communication, INTERSPEECH 2018 ; Conference date: 02-09-2018 Through 06-09-2018",

year = "2018",

doi = "10.21437/interspeech.2018-1159",

language = "אנגלית",

volume = "2018-September",

pages = "2868--2872",

journal = "Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",

issn = "2308-457X",

}

TY - JOUR

T1 - Word emphasis prediction for expressive text to speech

AU - Mass, Yosi

AU - Shechtman, Slava

AU - Mordechay, Moran

AU - Hoory, Ron

AU - Shalom, Oren Sar

AU - Lev, Guy

AU - Konopnicki, David

PY - 2018

Y1 - 2018

N2 - Word emphasis prediction is an important part of expressive prosody generation in modern Text-To-Speech (TTS) systems. We present a method for predicting emphasized words for expressive TTS, based on a Deep Neural Network (DNN). We show that the presented method outperforms machine learning methods based on hand-crafted features in terms of objective metrics such as precision and recall. Using a listening test, we further demonstrate that the contribution of the predicted emphasized words to the expressiveness of the synthesized speech is subjectively perceivable.

AB - Word emphasis prediction is an important part of expressive prosody generation in modern Text-To-Speech (TTS) systems. We present a method for predicting emphasized words for expressive TTS, based on a Deep Neural Network (DNN). We show that the presented method outperforms machine learning methods based on hand-crafted features in terms of objective metrics such as precision and recall. Using a listening test, we further demonstrate that the contribution of the predicted emphasized words to the expressiveness of the synthesized speech is subjectively perceivable.

KW - Deep learning

KW - Expressive text to speech

KW - Prosody

KW - Speech synthesis

KW - Word emphasis

UR - http://www.scopus.com/inward/record.url?scp=85055010379&partnerID=8YFLogxK

U2 - 10.21437/interspeech.2018-1159

DO - 10.21437/interspeech.2018-1159

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.conferencearticle???

AN - SCOPUS:85055010379

SN - 2308-457X

VL - 2018-September

SP - 2868

EP - 2872

JO - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

JF - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

T2 - 19th Annual Conference of the International Speech Communication, INTERSPEECH 2018

Y2 - 2 September 2018 through 6 September 2018

ER -

Word emphasis prediction for expressive text to speech

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this