Unified Approach to Development of ASR Systems for East Slavic Languages

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F46747885%3A24220%2F17%3A00004820" target="_blank" >RIV/46747885:24220/17:00004820 - isvavai.cz</a>
Result on the web
<a href="http://dx.doi.org/10.1007/978-3-319-68456-7_16" target="_blank" >http://dx.doi.org/10.1007/978-3-319-68456-7_16</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1007/978-3-319-68456-7_16" target="_blank" >10.1007/978-3-319-68456-7_16</a>

Result language
angličtina
Original language name
Unified Approach to Development of ASR Systems for East Slavic Languages
Original language description
This paper deals with the development of language specific modules (lexicons, phonetic inventories, LMs and AMs) for Russian, Ukrainian and Belarusian (used by 260M, 45M and 3M native speakers, respectively). Instead of working on each language separately, we adopt a common approach that allows us to share data and tools, yet taking into account language unique features. We utilize only freely available text and audio data that can be found on web pages of major newspaper and broadcast publishers. This must be done with large care, as the 3 languages are often mixed in spoken and written media. So, one component of the automated training process is a language identification module. At the output of the complete process there are 3 pronunciation lexicons (each about 300K words), 3 partly shared phoneme sets, and corresponding acoustic (DNN) and language (N-gram) models. We employ them in our media monitoring system and provide results achieved on a test set made of several complete TV news in all the 3 languages. The WER values vary in range from 24 to 36%.
Czech name
—
Czech description
—

Project
<a href="/en/project/TA04010199" target="_blank" >TA04010199: MULTILINMEDIA - Multilingual Multimedia Monitoring and Analyzing Platform</a><br>
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year
2017
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISBN
9783319684550
ISSN
0302-9743
e-ISSN
—
Number of pages
11
Pages from-to
193-203
Publisher name
Springer Verlag
Place of publication
Německo
Event location
Le Mans, Francie
Event date
Jan 1, 2017
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—

Similar results(10)