This HTML5 document contains 41 embedded RDF statements represented using HTML+Microdata notation.

The embedded RDF content will be recognized by any processor of HTML5 Microdata.

Namespace Prefixes

Prefix	IRI
dcterms	http://purl.org/dc/terms/
n15	http://localhost/temp/predkladatel/
n16	http://linked.opendata.cz/resource/domain/vavai/projekt/
n10	http://linked.opendata.cz/resource/domain/vavai/riv/tvurce/
n17	http://linked.opendata.cz/resource/domain/vavai/vysledek/RIV%2F00216305%3A26230%2F07%3APU70818%21RIV08-GA0-26230___/
n13	http://linked.opendata.cz/ontology/domain/vavai/
n11	http://linked.opendata.cz/ontology/domain/vavai/riv/kodPristupu/
rdfs	http://www.w3.org/2000/01/rdf-schema#
skos	http://www.w3.org/2004/02/skos/core#
n3	http://linked.opendata.cz/ontology/domain/vavai/riv/
n2	http://linked.opendata.cz/resource/domain/vavai/vysledek/
rdf	http://www.w3.org/1999/02/22-rdf-syntax-ns#
n7	http://linked.opendata.cz/ontology/domain/vavai/riv/klicoveSlovo/
n9	http://linked.opendata.cz/ontology/domain/vavai/riv/duvernostUdaju/
xsdh	http://www.w3.org/2001/XMLSchema#
n19	http://linked.opendata.cz/ontology/domain/vavai/riv/jazykVysledku/
n4	http://linked.opendata.cz/ontology/domain/vavai/riv/aktivita/
n18	http://linked.opendata.cz/ontology/domain/vavai/riv/druhVysledku/
n14	http://linked.opendata.cz/ontology/domain/vavai/riv/obor/
n8	http://reference.data.gov.uk/id/gregorian-year/

Statements

Subject Item

n2:RIV%2F00216305%3A26230%2F07%3APU70818%21RIV08-GA0-26230___

rdf:type

n13:Vysledek skos:Concept

rdfs:seeAlso

http://www.fit.vutbr.cz/~grezl/publi/mlmi2007.pdf

dcterms:description

Byly prezentováný různé topologie neuronových sítí pro odhad parametrů pro rozpoznávání řeči. Neuronová sít s bottle-neckem byla zavedena do struktury zvané &quot;rozdelený kontext&quot;. Cílem bylo zmenšit velikost výsledné sítě, která slouží pro odhad příznaků. Když jsou bottle-neckové výstupy použity také jako finální výstupy z neuronové sítě, je dosažené i snížení chybovosti rozpoznávače.<br> This poster overviewthe newly proposed bottle-neck features and then examines the possibility of use of meural net structure with<br> bottle-neck in hierarchical neural net classifier such as Split<br> Context classifier.<br> <br> First, the neural net with bottle-neck is used in place of merger to<br> see whether the advantage seeen for single neural net will hold also<br> for hierarchical classifier. Then we use the bottle-neck neural nets<br> in place of context classifiers, using bottle-neck outputs as input to<br> a merger classifier. Finally, bottle-neck neural nets are used in both<br> stages of Split Context classifier. This improved Split Context<br> structure gains several advantages: The use of bottle-neck imply<br> size reduction of resulting classifier. Also, processing of classifier<br> output is smaller compare to probabilistic features. The WER reduction was achieved too.<br> <br> This poster overviewthe newly proposed bottle-neck features and then examines the possibility of use of meural net structure with<br> bottle-neck in hierarchical neural net classifier such as Split<br> Context classifier.<br> <br> First, the neural net with bottle-neck is used in place of merger to<br> see whether the advantage seeen for single neural net will hold also<br> for hierarchical classifier. Then we use the bottle-neck neural nets<br> in place of context classifiers, using bottle-neck outputs as input to<br> a merger classifier. Finally, bottle-neck neural nets are used in both<br> stages of Split Context classifier. This improved Split Context<br> structure gains several advantages: The use of bottle-neck imply<br> size reduction of resulting classifier. Also, processing of classifier<br> output is smaller compare to probabilistic features. The WER reduction was achieved too.<br> <br>

dcterms:title

Neural network topologies and bottle neck features in speech recognition Topologie neuronových sítí a bottle-neckové parametry v rozpoznávání řeči Neural network topologies and bottle neck features in speech recognition

skos:prefLabel

Neural network topologies and bottle neck features in speech recognition Neural network topologies and bottle neck features in speech recognition Topologie neuronových sítí a bottle-neckové parametry v rozpoznávání řeči

skos:notation

RIV/00216305:26230/07:PU70818!RIV08-GA0-26230___

n3:aktivita

n4:P

n3:aktivity

P(GA102/05/0278)

n3:dodaniDat

n8:2008

n3:domaciTvurceVysledku

n10:8304874 n10:8912416 n10:4495896

n3:druhVysledku

n18:A

n3:duvernostUdaju

n9:S

n3:entitaPredkladatele

n17:predkladatel

n3:idSjednocenehoVysledku

436986

n3:idVysledku

RIV/00216305:26230/07:PU70818

n3:jazykVysledku

n19:eng

n3:klicovaSlova

neural networks, topologies, speech recognition, bottle-neck features<br>

n3:klicoveSlovo

n7:topologies n7:bottle-neck%20features%3Cbr%3E n7:speech%20recognition n7:neural%20networks

n3:kodPristupu

n11:V

n3:kontrolniKodProRIV

[940C89371228]

n3:obor

n14:JC

n3:pocetDomacichTvurcuVysledku

3

n3:pocetTvurcuVysledku

3

n3:projekt

n16:GA102%2F05%2F0278

n3:rokUplatneniVysledku

n8:2007

n3:tvurceVysledku

Grézl, František Černocký, Jan Karafiát, Martin

n15:organizacniJednotka

26230