1 |
Analyzing the impact of speaker localization errors on speech separation for automatic speech recognition
|
|
|
|
In: EUSIPCO 2020 - 28th European Signal Processing Conference ; https://hal.inria.fr/hal-02355669 ; EUSIPCO 2020 - 28th European Signal Processing Conference, Jan 2021, Amsterdam / Virtual, Netherlands. ⟨10.23919/Eusipco47968.2020.9287541⟩ ; https://eusipco2020.org/ (2021)
|
|
BASE
|
|
Show details
|
|
2 |
Orofacial somatosensory inputs change the auditory categorization in perceptual adaptation training
|
|
|
|
In: 13th Annual Meeting of the Society for the Neurobiology of Language ; https://hal.archives-ouvertes.fr/hal-03460304 ; 13th Annual Meeting of the Society for the Neurobiology of Language, Oct 2021, Virtual, United States (2021)
|
|
BASE
|
|
Show details
|
|
3 |
First Align, then Predict: Understanding the Cross-Lingual Ability of Multilingual BERT
|
|
|
|
In: https://hal.inria.fr/hal-03161685 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
4 |
First Align, then Predict: Understanding the Cross-Lingual Ability of Multilingual BERT
|
|
|
|
In: EACL 2021 - The 16th Conference of the European Chapter of the Association for Computational Linguistics ; https://hal.inria.fr/hal-03239087 ; EACL 2021 - The 16th Conference of the European Chapter of the Association for Computational Linguistics, Apr 2021, Kyiv / Virtual, Ukraine ; https://2021.eacl.org/ (2021)
|
|
Abstract:
International audience ; Multilingual pretrained language models have demonstrated remarkable zero-shot crosslingual transfer capabilities. Such transfer emerges by fine-tuning on a task of interest in one language and evaluating on a distinct language, not seen during the fine-tuning. Despite promising results, we still lack a proper understanding of the source of this transfer. Using a novel layer ablation technique and analyses of the model's internal representations, we show that multilingual BERT, a popular multilingual language model, can be viewed as the stacking of two sub-networks: a multilingual encoder followed by a taskspecific language-agnostic predictor. While the encoder is crucial for cross-lingual transfer and remains mostly unchanged during finetuning, the task predictor has little importance on the transfer and can be reinitialized during fine-tuning. We present extensive experiments with three distinct tasks, seventeen typologically diverse languages and multiple domains to support our hypothesis.
|
|
Keyword:
[INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
|
|
URL: https://hal.inria.fr/hal-03239087 https://hal.inria.fr/hal-03239087/file/2021.eacl-main.189.pdf https://hal.inria.fr/hal-03239087/document
|
|
BASE
|
|
Hide details
|
|
5 |
Language recognition on unknown conditions: the LORIA-Inria-MULTISPEECH system for AP20-OLR Challenge
|
|
|
|
In: Interspeech ; https://hal.archives-ouvertes.fr/hal-03228823 ; Interspeech, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
6 |
Language recognition on unknown conditions: the LORIA-Inria-MULTISPEECH system for AP20-OLR Challenge
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03228823 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
7 |
Saliva Pepsin Concentration of Laryngopharyngeal Reflux Patients Is Influenced by Meals Consumed Before the Samples
|
|
|
|
In: ISSN: 0023-852X ; Laryngoscope ; https://hal.archives-ouvertes.fr/hal-03141469 ; Laryngoscope, Wiley, 2021, 131 (2), pp.350-359. ⟨10.1002/lary.28756⟩ (2021)
|
|
BASE
|
|
Show details
|
|
8 |
Frameless fictions: embodiment, affect, and unruly encounters in VR and virtual environments
|
|
|
|
BASE
|
|
Show details
|
|
9 |
What do disyllabic words tell us about syllable structure, vowel quality, and stress in English?
|
|
|
|
BASE
|
|
Show details
|
|
11 |
Priorities, visions, identities, and normative approaches: a corpus-driven discourse analysis of the official discourse between China and the EU (1994-2019)
|
|
|
|
BASE
|
|
Show details
|
|
12 |
Edinburgh Cognitive and Behavioural ALS Screen (ECAS) in the assessment of early onset dementia
|
|
|
|
BASE
|
|
Show details
|
|
13 |
Quo vadis? Epigraphy, language, and people in the making of Roman Italy
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Becoming woman in the land of women: investigating the paradigm of the individual versus the collective in contemporary feminist utopianism
|
|
|
|
BASE
|
|
Show details
|
|
16 |
Bilingualism meets autism: an investigation of executive functions profiles in English-Arabic children
|
|
|
|
BASE
|
|
Show details
|
|
17 |
RNN Language Model Estimation for Out-of-Vocabulary Words
|
|
|
|
In: Lecture Notes in Artificial Intelligence ; https://hal.archives-ouvertes.fr/hal-03054936 ; Lecture Notes in Artificial Intelligence, Springer, In press, 12598, ⟨10.1007/978-3-030-66527-2_15⟩ (2020)
|
|
BASE
|
|
Show details
|
|
19 |
L2 writers referencing corpora to address accuracy: a qualitative analysis of learners' lexicogrammatical error corrections
|
|
|
|
BASE
|
|
Show details
|
|
20 |
(S)mothered in translation? (Re)translating the female Bildungsroman in the twentieth century in English and French
|
|
|
|
BASE
|
|
Show details
|
|
|
|