Vložit kód Přihlásit se
Technická podpora
COLIPS

The Chinese and Oriental Languages Information Processing Society, or COLIPS in short, is a non-profit professional organisation that was established in 1988 to advance the research of Chinese and other Asian languages. It promotes the free exchange of information about information processing of these languages in the best scientific and professional tradition. COLIPS organizes international conferences, short courses and seminars for members and the public. COLIPS publishes the International Journal of Asian Languages Processing four times a year that is circulated world-wide. Having its members from all over the world, COLIPS is based in Singapore. It is one of the founding members of Asian Federation of Natural Language Processing (AFNLP).

Konference:

Fusing Language Information from Diverse Data Sources for Phonotactic Language Recognition

Rok: 2012

The baseline approach in building phonotactic language recognition systems is to characterize each language by a single phonotactic model generated from all the available languagespecific training data. When several data sources are available for a given target language, system performance can be improved using language source-dependent phonotactic models. In this case, the common practice is to fuse language source information (i.e., the phonotactic scores for each language/ source) early (at the input) to the backend. This paper proposes to postpone the fusion to the end (at the output) of the backend. In this case, the language recognition score can be estimated from well-calibrated language source scores. Experiments were conducted using the NIST LRE 2007 and the NIST LRE 2009 evaluation data sets with the 30s condition. On the NIST LRE 2007 eval data, a Cavg of 0.9% is obtained for the closed-set task and 2.5% for the open-set task. Compared to the common practice of early fusion, these results represent relative improvements of 18% and 11%, for the closed-set and open-set tasks, respectively. Initial tests on the NIST LRE 2009 eval data gave no improvement on the closedset task. Moreover, the Cllr measure indicates that language recognition scores estimated by the proposed approach are better calibrated than the common practice (early fusion).



agentura Motiv P s.r.o.
Řehořova 726/14, 618 00 Brno

E-mail: helpdesk@motivp.com
Tel: +420 545 234 698
www.motivp.com
Copyright © 1996-2017, agentura Motiv P s.r.o. Všechna práva vyhrazena.


Agentura Motiv P používá na svém webu www.MotivP.com soubory cookie. Používáním webu vyjadřujete souhlas.