This HTML5 document contains 30 embedded RDF statements represented using HTML+Microdata notation.

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

Namespace Prefixes

PrefixIRI
n11https://kar.kent.ac.uk/69101/
dctermshttp://purl.org/dc/terms/
n2https://kar.kent.ac.uk/id/eprint/
wdrshttp://www.w3.org/2007/05/powder-s#
n21http://purl.org/ontology/bibo/status/
dchttp://purl.org/dc/elements/1.1/
n13https://kar.kent.ac.uk/id/subject/
rdfshttp://www.w3.org/2000/01/rdf-schema#
n19doi:10.1109/
n15https://demo.openlinksw.com/about/id/entity/https/raw.githubusercontent.com/annajordanous/CO644Files/main/
n6http://eprints.org/ontology/
n16https://kar.kent.ac.uk/id/event/
bibohttp://purl.org/ontology/bibo/
n5https://kar.kent.ac.uk/id/org/
rdfhttp://www.w3.org/1999/02/22-rdf-syntax-ns#
owlhttp://www.w3.org/2002/07/owl#
n7https://kar.kent.ac.uk/id/document/
n8https://kar.kent.ac.uk/id/
xsdhhttp://www.w3.org/2001/XMLSchema#
n20https://kar.kent.ac.uk/id/eprint/69101#
n17https://demo.openlinksw.com/about/id/entity/https/www.cs.kent.ac.uk/people/staff/akj22/materials/CO644/
n4https://kar.kent.ac.uk/id/person/

Statements

Subject Item
n2:69101
rdf:type
n6:EPrint bibo:AcademicArticle bibo:Article n6:ConferenceItemEPrint
rdfs:seeAlso
n11:
owl:sameAs
n19:SLT.2018.8639522
n6:hasAccepted
n7:3060865
n6:hasDocument
n7:3060865 n7:3060866 n7:3092451 n7:3092452 n7:3092449 n7:3092450
dc:hasVersion
n7:3060865
dcterms:title
Improved Conditional Generative Adversarial Net Classification For Spoken Language Recognition
wdrs:describedby
n15:export_kar_RDFN3.n3 n17:export_kar_RDFN3.n3
dcterms:date
2018-12-18
dcterms:creator
n4:ext-fd53534108ea97c5c295193085a0d7f5 n4:ext-xm39@kent.ac.uk n4:ext-i.v.mcloughlin@kent.ac.uk n4:ext-d95cd0c6cc7745a518c19ef34236df70
bibo:status
n21:peerReviewed n21:published
dcterms:publisher
n5:ext-af0a9a5baed87c407844a3f5db44597c
bibo:abstract
Recent research on generative adversarial nets (GAN) for language identification (LID) has shown promising results. In this paper, we further exploit the latent abilities of GAN networks to firstly combine them with deep neural network (DNN)-based i-vector approaches and then to improve the LID model using conditional generative adversarial net (cGAN) classification. First, phoneme dependent deep bottleneck features (DBF) combined with output posteriors of a pre-trained DNN for automatic speech recognition (ASR) are used to extract i-vectors in the normal way. These i-vectors are then classified using cGAN, and we show an effective method within the cGAN to optimize parameters by combining both language identification and verification signals as supervision. Results show firstly that cGAN methods can significantly outperform DBF DNN i-vector methods where 49-dimensional i-vectors are used, but not where 600-dimensional vectors are used. Secondly, training a cGAN discriminator network for direct classification has further benefit for low dimensional i-vectors as well as short utterances with high dimensional i-vectors. However, incorporating a dedicated discriminator network output layer for classification and optimizing both classification and verification loss brings benefits in all test cases.
dcterms:isPartOf
n8:repository
dcterms:subject
n13:T
bibo:authorList
n20:authors
bibo:presentedAt
n16:ext-dadb48f93f99dbfcd9c8ae0e0ac9699d