[ChimeText] 5 May (tomorrow!) Gina-Anne Levow (U. Manchester) / Context and Learning in Multilingual Tone and Pitch Accent Recognition

[Min-Yen Kan]

A reminder of tomorrow's seminar by Gina-Anne Levow, from the
University of Manchester.

TITLE: Context and Learning in Multilingual Tone and Pitch Accent Recognition
SPEAKER: Gina-Anne Levow / University of Manchester
DATE: Tuesday, 5 May 2009
TIME: 10am-11am
VENUE: MR6 (AS6 #05-10)

Chaired by A/P Ng Hwee Tou

ABSTRACT:

Tone and intonation play a crucial role across many languages.
However, the use and structure of tone varies widely, ranging from
lexical tone which determines word identity to pitch accent signalling
information status.   In this work, we employ a uniform representation
of acoustic features for recognition of Mandarin tone, isiZulu tone,
and English pitch accent. The representation captures both local tone
height and shape as well as contextual coarticulatory and phrasal
influences.

By exploiting  multiclass Support Vector Machines as a discriminative
classifier, we achieve competitive rates of tone and pitch accent
recognition. We  demonstrate the greater importance of modeling
preceding local context, which yields up to 24% reduction in error
over modeling the following context, and further demonstrate that
alternate acoustic features such band energy can improve tone
recognition, for challenging cases such as neutral tone.

While these approaches to this recognition task have relied upon fully
supervised learning methods employing extensive collections of
manually tagged data obtained at substantial time and financial cost,
we next explore two approaches to tone learning with substantial
reductions in training data.  We employ both unsupervised clustering
and semi-supervised learning to recognize pitch accent and tone, based
on the  intrinsic structure of the tones in acoustic space.  In
unsupervised tone and pitch accent clustering experiments, we achieve
75% to 96% of accuracy rates achieved with large training data sets.
For semi-supervised training with only small numbers of labeled
examples, accuracies reach 90-98% of levels obtained with hundreds or
thousands of labeled examples.  These results indicate that the
intrinsic structure of tone and pitch accent acoustics can be
exploited to reduce the need for costly labeled training data for tone
learning and recognition.

BIODATA:

Gina-Anne Levow is a Research Fellow at the National Centre for Text
Mining in the School of Computer Science at the University of
Manchester. From 2001 to 2008, she served as an Assistant Professor in
the Computer Science Department at the University of Chicago, where
she still holds an appointment as Research Associate (Assistant
Professor).  She received undergraduate degrees in Computer Science
and Oriental Studies from the University of Pennsylvania  and her
Master's and Ph.D. from Massachusetts Institute of Technology.  Her
research is strongly multi-lingual and spans natural language
processing, spoken language processing, and information retrieval.
Specific areas of interest include prosody in discourse and dialogue,
spoken and multi-lingual document retrieval, multi-modal interaction,
and entity and event extraction.

Upcoming Talks:
5 May - Gina-Anne Levow / Context and Learning in Multilingual Tone
and Pitch Accent Recognition
30 Jun - Wang Kai and Ye Shiren / "A Syntactic Tree Matching Approach
to Finding Similar Questions in Community-based QA Services" and
"Summarizing Definition from Wikipedia"