FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2340254635> ?p ?o ?g. }

• W2340254635 abstract "AbstractManual transcription of audio databases for the development of automatic speech recognition (ASR) systemsis a costly and time-consuming process. In the context of deriving acoustic models adapted to a speci capplication, or in low-resource scenarios, it is therefore essential to explore alternatives capable of improvingspeech recognition results. In this paper, we investigate the relevance of foreign data characteristics, inparticular domain and language, when using this data as an auxiliary data source for training ASR acousticmodels based on deep neural networks (DNNs). The acoustic models are evaluated on a challenging bilingualdatabase within the scope of the MediaParl project. Experimental results suggest that in-language (butout-of-domain) data is more bene cial than in-domain (but out-of-language) data when employed in eithersupervised or semi-supervised training of DNNs. The best performing ASR system, an HMM/GMM acousticmodel that exploits DNN as a discriminatively trained feature extractor outperforms the best performingHMM/DNN hybrid by about 5% relative (in terms of WER). An accumulated relative gain with respect to theMFCC-HMM/GMM baseline is about 30% WER.Keywords: Automatic speech recognition; Deep learning for speech; Acoustic model adaptation;Semi-supervised training" @default.
• W2340254635 created "2016-06-24" @default.
• W2340254635 creator A5037750159 @default.
• W2340254635 creator A5039529498 @default.
• W2340254635 creator A5052265530 @default.
• W2340254635 creator A5076409146 @default.
• W2340254635 creator A5081808619 @default.
• W2340254635 date "2015-01-01" @default.
• W2340254635 modified "2023-09-27" @default.
• W2340254635 title "EXPLOITING FOREIGN RESOURCES FOR" @default.
• W2340254635 cites W156213964 @default.
• W2340254635 cites W1607988020 @default.
• W2340254635 cites W162588823 @default.
• W2340254635 cites W1738772085 @default.
• W2340254635 cites W184563473 @default.
• W2340254635 cites W1966812932 @default.
• W2340254635 cites W1970890968 @default.
• W2340254635 cites W197277712 @default.
• W2340254635 cites W1988596541 @default.
• W2340254635 cites W1993660824 @default.
• W2340254635 cites W1993952617 @default.
• W2340254635 cites W1994606281 @default.
• W2340254635 cites W1996535874 @default.
• W2340254635 cites W2002342963 @default.
• W2340254635 cites W2012190582 @default.
• W2340254635 cites W2025198378 @default.
• W2340254635 cites W2043661469 @default.
• W2340254635 cites W2049123401 @default.
• W2340254635 cites W2056825827 @default.
• W2340254635 cites W2090764203 @default.
• W2340254635 cites W2094147890 @default.
• W2340254635 cites W2097491464 @default.
• W2340254635 cites W2105594594 @default.
• W2340254635 cites W2112676096 @default.
• W2340254635 cites W2120209245 @default.
• W2340254635 cites W2124558353 @default.
• W2340254635 cites W2125234026 @default.
• W2340254635 cites W2125610823 @default.
• W2340254635 cites W2125838338 @default.
• W2340254635 cites W2127982613 @default.
• W2340254635 cites W2131342762 @default.
• W2340254635 cites W2136922672 @default.
• W2340254635 cites W2137222873 @default.
• W2340254635 cites W2138857742 @default.
• W2340254635 cites W2146168991 @default.
• W2340254635 cites W2147590749 @default.
• W2340254635 cites W2147768505 @default.
• W2340254635 cites W2150907703 @default.
• W2340254635 cites W2152483743 @default.
• W2340254635 cites W2158069733 @default.
• W2340254635 cites W2160614052 @default.
• W2340254635 cites W2165357419 @default.
• W2340254635 cites W2165712214 @default.
• W2340254635 cites W2169189000 @default.
• W2340254635 cites W217970951 @default.
• W2340254635 cites W2189391786 @default.
• W2340254635 cites W22168010 @default.
• W2340254635 cites W2405962136 @default.
• W2340254635 cites W2407897255 @default.
• W2340254635 cites W2616139854 @default.
• W2340254635 cites W44815768 @default.
• W2340254635 cites W64472890 @default.
• W2340254635 cites W2339821021 @default.
• W2340254635 hasPublicationYear "2015" @default.
• W2340254635 type Work @default.
• W2340254635 sameAs 2340254635 @default.
• W2340254635 citedByCount "0" @default.
• W2340254635 crossrefType "journal-article" @default.
• W2340254635 hasAuthorship W2340254635A5037750159 @default.
• W2340254635 hasAuthorship W2340254635A5039529498 @default.
• W2340254635 hasAuthorship W2340254635A5052265530 @default.
• W2340254635 hasAuthorship W2340254635A5076409146 @default.
• W2340254635 hasAuthorship W2340254635A5081808619 @default.
• W2340254635 hasConcept C134306372 @default.
• W2340254635 hasConcept C137293760 @default.
• W2340254635 hasConcept C138885662 @default.
• W2340254635 hasConcept C151730666 @default.
• W2340254635 hasConcept C154945302 @default.
• W2340254635 hasConcept C155635449 @default.
• W2340254635 hasConcept C158154518 @default.
• W2340254635 hasConcept C165696696 @default.
• W2340254635 hasConcept C17744445 @default.
• W2340254635 hasConcept C199539241 @default.
• W2340254635 hasConcept C23224414 @default.
• W2340254635 hasConcept C2776401178 @default.
• W2340254635 hasConcept C2779343474 @default.
• W2340254635 hasConcept C28490314 @default.
• W2340254635 hasConcept C33923547 @default.
• W2340254635 hasConcept C36503486 @default.
• W2340254635 hasConcept C38652104 @default.
• W2340254635 hasConcept C41008148 @default.
• W2340254635 hasConcept C41895202 @default.
• W2340254635 hasConcept C50644808 @default.
• W2340254635 hasConcept C61328038 @default.
• W2340254635 hasConcept C86803240 @default.
• W2340254635 hasConceptScore W2340254635C134306372 @default.
• W2340254635 hasConceptScore W2340254635C137293760 @default.
• W2340254635 hasConceptScore W2340254635C138885662 @default.
• W2340254635 hasConceptScore W2340254635C151730666 @default.
• W2340254635 hasConceptScore W2340254635C154945302 @default.

• next