Conversational speech is characterized by prosodic variability which makes pitch accent prediction for this genre especially difficult. The linguistic literature points out that complex features such as information status, contrast and animacy help predict pitch accent placement. In this paper, we use a corpus annotated for such features to determine if they improve prominence prediction over traditional shallow features such as frequency and part-of-speech, or over new ones that we introduce. We demonstrate that while correlated with prominence, complex linguistic features do not improve prediction accuracy. Furthermore, the performance of our classifier is quite close to the ceiling defined by variability in human accent placement. An oracle experiment demonstrates, though, that at least some accuracy improvement is still possible.
%0 Conference Paper
%1 brenier06nonutility
%A Brenier, Jason M.
%A Nenkova, Ani
%A Kothari, Anubha
%A Whitton, Laura
%A Beaver, David
%A Jurafsky, Dan
%B Proceedings of the IEEE / ACL 2006 Workshop on Spoken Language Technology
%D 2006
%K 2006 NT2OD nlp parser parsetree stanford
%T The (Non)Utility of Linguistic Features for Predicting Prominence in Spontaneous Speech
%U http://nlp.stanford.edu/pubs/1086_brenier.pdf
%X Conversational speech is characterized by prosodic variability which makes pitch accent prediction for this genre especially difficult. The linguistic literature points out that complex features such as information status, contrast and animacy help predict pitch accent placement. In this paper, we use a corpus annotated for such features to determine if they improve prominence prediction over traditional shallow features such as frequency and part-of-speech, or over new ones that we introduce. We demonstrate that while correlated with prominence, complex linguistic features do not improve prediction accuracy. Furthermore, the performance of our classifier is quite close to the ceiling defined by variability in human accent placement. An oracle experiment demonstrates, though, that at least some accuracy improvement is still possible.
@inproceedings{brenier06nonutility,
abstract = {Conversational speech is characterized by prosodic variability which makes pitch accent prediction for this genre especially difficult. The linguistic literature points out that complex features such as information status, contrast and animacy help predict pitch accent placement. In this paper, we use a corpus annotated for such features to determine if they improve prominence prediction over traditional shallow features such as frequency and part-of-speech, or over new ones that we introduce. We demonstrate that while correlated with prominence, complex linguistic features do not improve prediction accuracy. Furthermore, the performance of our classifier is quite close to the ceiling defined by variability in human accent placement. An oracle experiment demonstrates, though, that at least some accuracy improvement is still possible.},
added-at = {2007-03-25T17:32:18.000+0200},
author = {Brenier, Jason M. and Nenkova, Ani and Kothari, Anubha and Whitton, Laura and Beaver, David and Jurafsky, Dan},
biburl = {https://www.bibsonomy.org/bibtex/2086fd2c357b0e302cc4df16b96eadc1c/butonic},
booktitle = {Proceedings of the IEEE / ACL 2006 Workshop on Spoken Language Technology},
interhash = {c855d2edc887be08176374100263edca},
intrahash = {086fd2c357b0e302cc4df16b96eadc1c},
keywords = {2006 NT2OD nlp parser parsetree stanford},
organization = {The Stanford Natural Language Processing Group},
school = {Stanford University},
timestamp = {2007-03-25T17:32:18.000+0200},
title = {The ({N}on){U}tility of {L}inguistic {F}eatures for {P}redicting {P}rominence in {S}pontaneous {S}peech},
url = {http://nlp.stanford.edu/pubs/1086_brenier.pdf},
year = 2006
}