Mostra el registre d'ítem simple

dc.contributorRuiz Costa-Jussà, Marta
dc.contributorGallego Olsina, Gerard Ion
dc.contributor.authorAlastruey Lasheras, Belén
dc.contributor.otherUniversitat Politècnica de Catalunya. Departament de Ciències de la Computació
dc.date.accessioned2021-07-14T12:19:19Z
dc.date.available2021-07-14T12:19:19Z
dc.date.issued2021-07
dc.identifier.urihttp://hdl.handle.net/2117/349294
dc.description.abstractIn this thesis, we propose a new approach for Speech-to-Text translation, where thanks to an efficient Transformer we can work with a spectrogram without having to use convolutional layers before the Transformer. This allows the encoder to learn directly from the spectrogram and no information is lost, which we believe could be profitable. We have created an encoder-decoder model, where the encoder is an efficient Transformer -the Longformer- and the decoder is a traditional Transformer decoder. Firstly we trained our model for an Automatic Speech Recognition (ASR) task, and then for Speech Translation using the ASR pre-trained encoder. Our results are close to the ones obtained with convolutional layers and a regular Transformer, showing less than a 10% relative reduction of the performance, meaning that this is a great starting point for a promising research path.
dc.language.isoeng
dc.publisherUniversitat Politècnica de Catalunya
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/3.0/es/
dc.subjectÀrees temàtiques de la UPC::Matemàtiques i estadística
dc.subjectÀrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
dc.subject.lcshArtificial intelligence
dc.subject.otherDeep Learning
dc.subject.otherTransformer
dc.subject.otherSpeech-to-Text
dc.subject.otherSpeech Translation
dc.subject.otherMachine Translation
dc.subject.otherNeural Network
dc.subject.otherAutomatic Speech Recognition
dc.titleEfficient transformers for direct speech translation
dc.typeBachelor thesis
dc.subject.lemacIntel·ligència artificial
dc.subject.amsClassificació AMS::68 Computer science::68T Artificial intelligence
dc.identifier.slugFME-2152
dc.rights.accessOpen Access
dc.date.updated2021-07-14T05:22:35Z
dc.audience.educationlevelGrau
dc.audience.mediatorUniversitat Politècnica de Catalunya. Facultat de Matemàtiques i Estadística
dc.audience.degreeGRAU EN MATEMÀTIQUES (Pla 2009)


Fitxers d'aquest items

Thumbnail

Aquest ítem apareix a les col·leccions següents

Mostra el registre d'ítem simple