LINGUIST List 19.2623
|
Tue Aug 26 2008
Confs: Phonology/India
Editor for this issue: Ania Kubisz
<ania linguistlist.org>
|
To post to LINGUIST, use our convenient web form at
http://linguistlist.org/LL/posttolinguist.html.
|
Directory
1. Kalika
Bali,
Consonant Challenge for Indian Languages
Message 1: Consonant Challenge for Indian Languages
|
Date: 25-Aug-2008
From: Kalika Bali <kalikabmicrosoft.com>
Subject: Consonant Challenge for Indian Languages
 E-mail this message to a friend
Consonant Challenge for Indian Languages
Short Title: CCIL
Date: 15-Dec-2008 - 18-Dec-2008
Location: Goa, India
Contact: Kalika Bali
Contact Email: kalikabmicrosoft.com
Meeting URL: http://ragashri.ee.iisc.ernet.in/ILCC
Linguistic Field(s): Phonology
Meeting Description
Consonant Challenge for Indian Languages
Detection and Recognition of Consonants in Indian Language Speech Data
(Special Session of SLT08 http://www.slt2008.org/default.asp)
Call for Participation
In order to promote speech technology research in Indian Languages and to better
understand any specific issues related to speech recognition of these languages
and the possible means to address them, we are pleased to announce a Consonant
Challenge in Indian Languages.
The task involves detection of consonants (in CV, VC, CVC and VCV positions) in
a surprise language. Training data is provided in 6 Indian languages, namely,
Assamese, Bengali, Hindi, Marathi, Tamil and Telugu to all registered
participants. Based on the recognition results received by the organizers and
evaluated by the program committee, the highest two accuracy results will be
awarded a cash prize of USD 500 and USD 250 respectively.
The results will be presented in a special session at SLT 08 in Goa, India.
Background:
Consonant detection in speech by a machine based on purely spectral features is
always problematic due to a number of reasons like the unvoiced (no-energy)
portions of stop consonants that can be confused with real silence, the high
energy fricative noise that maybe confused with environmental or additive noise,
and the vowel like spectrum of the liquids, the nasals and the semi-vowels that
make them hard to distinguish from vowels. This problem is further compounded in
Indian languages where the number of consonants can go from around 23 (in Tamil)
to almost 40 (in Hindi-Urdu). For example, acoustic phonetic features like voice
and aspiration form a four way contrast in many Indian language stop and
affricate consonants. Further, stop consonants occur for at least four, that is,
labial, dental, retroflex, and velar, place of articulation (this can go to 5 or
6 for some languages like Malayalam and Hindi-Urdu). Though all Indian Languages
come from four major language families (Indo-Aryan, Dravidian, Austronesian and
Tibeto-Burman, with the majority from the former two), the languages have
co-existed for a long time to have borrowed and shared features even at the
phonetic level. For example, the borrowing of retroflex sounds from Dravidian to
Indo-European and of aspiration as a feature of stops the other way around.
From a Speech Recognition perspective, a deeper understanding of how consonants
are detected and recognized can not only help us better understand how to model
these sounds (ref. difference between human and computer consonant recognition)
but also, in the specific case of Indian languages, open up research issues
into model adaptation from one language to another (related)language. This might
allow researchers to explore ways and means to scale from one language to
another where resources in terms of training data are limited
Register:
To register for the CCIL, please mail the organizing chairs by 25th August 2008.
Data will be released to the registered participants only
All participants for the CCIL will have to register for the main SLT08.
Important Dates
Registration for CCIL: 25th August 2008
Training Data release to registered participants: 29th August 2008
Test Data in surprise language made available: 8th September 2008
Organizing Committee:
Prof. AG Ramakrishanan, Dept of Electrical Engineering, Indian Institute of
Science, Bangalore
Kalika Bali, Microsoft Research India
Program Committee:
Prof. Hema Murthy, IIT Madras, Chennai
Prof. Preeti Rao, IIT Bombay, Mumbai
Dr. Mallikarjun, CIIL, Mysore
Prof. Roni Rosenfeld, CMU, Pittsburgh
Dr. Shyamal Das Mandal, CDAC Kolkata
Dr. Amitav Das, Microsoft Research
Dr. Ashish Verma, IBM IRL, New Delhi
Contact:
Please mail the organising chairs to register for the challenge at:
ramkiagee.iisc.ernet.in
kalikabmicrosoft.com
URL:
http://ragashri.ee.iisc.ernet.in/ILCC
Read more issues|LINGUIST home page|Top of issue
|
|
Please report any bad links or misclassified data
LINGUIST Homepage | Read
LINGUIST | Contact us
While the LINGUIST List makes every effort to ensure the linguistic relevance of sites listed
on its pages, it cannot vouch for their contents.
|
|
