Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks
International audience This work presents a novel Variational Neural Machine Translation (VNMT) architecture with enhanced robustness properties, which we investigate through a detailed case-study addressing noisy French user-generated content (UGC) translation to English. We show that the proposed...
Main Authors: | , , |
---|---|
Other Authors: | , , , , , , , |
Format: | Conference Object |
Language: | English |
Published: |
HAL CCSD
2023
|
Subjects: | |
Online Access: | https://hal.science/hal-04384748 https://hal.science/hal-04384748/document https://hal.science/hal-04384748/file/_Article__VNMT_nodalida.pdf |
id |
ftccsdartic:oai:HAL:hal-04384748v1 |
---|---|
record_format |
openpolar |
spelling |
ftccsdartic:oai:HAL:hal-04384748v1 2024-02-11T10:03:42+01:00 Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks Rosales Núñez, José, Carlos Seddah, Djamé Wisniewski, Guillaume Laboratoire Interdisciplinaire des Sciences du Numérique (LISN) Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS) Automatic Language Modelling and ANAlysis & Computational Humanities (ALMAnaCH) Inria de Paris Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria) Laboratoire de Linguistique Formelle (LLF - UMR7110) Centre National de la Recherche Scientifique (CNRS)-Université Paris Cité (UPCité) ANR-16-CE33-0021,PARSITI,Analyser l'impossible, Traduire l'improbable(2016) Torshavn, Faroe Islands 2023-05-22 https://hal.science/hal-04384748 https://hal.science/hal-04384748/document https://hal.science/hal-04384748/file/_Article__VNMT_nodalida.pdf en eng HAL CCSD hal-04384748 https://hal.science/hal-04384748 https://hal.science/hal-04384748/document https://hal.science/hal-04384748/file/_Article__VNMT_nodalida.pdf info:eu-repo/semantics/OpenAccess NoDaLiDa 2023 - 24th Nordic Conference on Computational Linguistics https://hal.science/hal-04384748 NoDaLiDa 2023 - 24th Nordic Conference on Computational Linguistics, May 2023, Torshavn, Faroe Islands Machine Translation User-generated content [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] info:eu-repo/semantics/conferenceObject Conference papers 2023 ftccsdartic 2024-01-20T23:46:20Z International audience This work presents a novel Variational Neural Machine Translation (VNMT) architecture with enhanced robustness properties, which we investigate through a detailed case-study addressing noisy French user-generated content (UGC) translation to English. We show that the proposed model, with results comparable or superior to state-of-the-art VNMT, improves performance over UGC translation in a zero-shot evaluation scenario while keeping optimal translation scores on in-domain test sets. We elaborate on such results by visualizing and explaining how neural learning representations behave when processing UGC noise. In addition, we show that VNMT enforces robustness to the learned embeddings, which can be later used for robust transfer learning approaches. Conference Object Faroe Islands Torshavn Archive ouverte HAL (Hyper Article en Ligne, CCSD - Centre pour la Communication Scientifique Directe) Faroe Islands |
institution |
Open Polar |
collection |
Archive ouverte HAL (Hyper Article en Ligne, CCSD - Centre pour la Communication Scientifique Directe) |
op_collection_id |
ftccsdartic |
language |
English |
topic |
Machine Translation User-generated content [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] |
spellingShingle |
Machine Translation User-generated content [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] Rosales Núñez, José, Carlos Seddah, Djamé Wisniewski, Guillaume Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
topic_facet |
Machine Translation User-generated content [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG] |
description |
International audience This work presents a novel Variational Neural Machine Translation (VNMT) architecture with enhanced robustness properties, which we investigate through a detailed case-study addressing noisy French user-generated content (UGC) translation to English. We show that the proposed model, with results comparable or superior to state-of-the-art VNMT, improves performance over UGC translation in a zero-shot evaluation scenario while keeping optimal translation scores on in-domain test sets. We elaborate on such results by visualizing and explaining how neural learning representations behave when processing UGC noise. In addition, we show that VNMT enforces robustness to the learned embeddings, which can be later used for robust transfer learning approaches. |
author2 |
Laboratoire Interdisciplinaire des Sciences du Numérique (LISN) Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS) Automatic Language Modelling and ANAlysis & Computational Humanities (ALMAnaCH) Inria de Paris Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria) Laboratoire de Linguistique Formelle (LLF - UMR7110) Centre National de la Recherche Scientifique (CNRS)-Université Paris Cité (UPCité) ANR-16-CE33-0021,PARSITI,Analyser l'impossible, Traduire l'improbable(2016) |
format |
Conference Object |
author |
Rosales Núñez, José, Carlos Seddah, Djamé Wisniewski, Guillaume |
author_facet |
Rosales Núñez, José, Carlos Seddah, Djamé Wisniewski, Guillaume |
author_sort |
Rosales Núñez, José, Carlos |
title |
Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
title_short |
Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
title_full |
Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
title_fullStr |
Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
title_full_unstemmed |
Multi-way Variational NMT for UGC: Improving Robustness in Zero-shot Scenarios via Mixture Density Networks |
title_sort |
multi-way variational nmt for ugc: improving robustness in zero-shot scenarios via mixture density networks |
publisher |
HAL CCSD |
publishDate |
2023 |
url |
https://hal.science/hal-04384748 https://hal.science/hal-04384748/document https://hal.science/hal-04384748/file/_Article__VNMT_nodalida.pdf |
op_coverage |
Torshavn, Faroe Islands |
geographic |
Faroe Islands |
geographic_facet |
Faroe Islands |
genre |
Faroe Islands Torshavn |
genre_facet |
Faroe Islands Torshavn |
op_source |
NoDaLiDa 2023 - 24th Nordic Conference on Computational Linguistics https://hal.science/hal-04384748 NoDaLiDa 2023 - 24th Nordic Conference on Computational Linguistics, May 2023, Torshavn, Faroe Islands |
op_relation |
hal-04384748 https://hal.science/hal-04384748 https://hal.science/hal-04384748/document https://hal.science/hal-04384748/file/_Article__VNMT_nodalida.pdf |
op_rights |
info:eu-repo/semantics/OpenAccess |
_version_ |
1790600007660863488 |