Recognition of radio broadcasted speech in a task with low perplexity and sparse training data

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F99%3A00041882" target="_blank" >RIV/49777513:23520/99:00041882 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Recognition of radio broadcasted speech in a task with low perplexity and sparse training data

Popis výsledku v původním jazyce

This paper describes an automatic recognition of a weather forecast transmitted by the Czech radio via the experimental speech recognition system being developed at the University of West Bohemia with support of the Johns Hopkins University. The experimental recognition system is designed as a speaker independent and is based on the HMM with mixture Gaussian continuous densities. A set of Czech phonological rules is used to provide phonetic transcription. Triphones and Czech phonetic decision trees areused to model a phonetic context. The system incorporates an n-gram language model. Two kinds of n-gram language model together with various smoothing techniques were tested to overcome very sparse trainingThe results of recognition experiments with twokinds of language models as well as with various language and acoustic model weights are presented in the paper.

Název v anglickém jazyce

Recognition of radio broadcasted speech in a task with low perplexity and sparse training data

Popis výsledku anglicky

This paper describes an automatic recognition of a weather forecast transmitted by the Czech radio via the experimental speech recognition system being developed at the University of West Bohemia with support of the Johns Hopkins University. The experimental recognition system is designed as a speaker independent and is based on the HMM with mixture Gaussian continuous densities. A set of Czech phonological rules is used to provide phonetic transcription. Triphones and Czech phonetic decision trees areused to model a phonetic context. The system incorporates an n-gram language model. Two kinds of n-gram language model together with various smoothing techniques were tested to overcome very sparse trainingThe results of recognition experiments with twokinds of language models as well as with various language and acoustic model weights are presented in the paper.

Druh

C - Kapitola v odborné knize

CEP obor

JD - Využití počítačů, robotika a její aplikace

OECD FORD obor

—

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>Z - Vyzkumny zamer (s odkazem do CEZ)

Rok uplatnění

1999

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název knihy nebo sborníku

Recognition of radio broadcasted speech in a task with low perplexity and sparse training data

ISBN

—

Počet stran výsledku

1

Strana od-do

—

Počet stran knihy

—

Název nakladatele

x

Místo vydání

X

Kód UT WoS kapitoly

—