Search

Your search keyword '"Melero, Maite"' showing total 45 results
Start Over Author "Melero, Maite" Publication Year Range Last 10 years
45 results on '"Melero, Maite"'

1. The power of Prompts: Evaluating and Mitigating Gender Bias in MT with LLMs

Author

Sant, Aleix, Escolano, Carlos, Mash, Audrey, Fornaciari, Francesca De Luca, and Melero, Maite

Subjects
Computer Science - Computation and Language
Abstract

This paper studies gender bias in machine translation through the lens of Large Language Models (LLMs). Four widely-used test sets are employed to benchmark various base LLMs, comparing their translation quality and gender bias against state-of-the-art Neural Machine Translation (NMT) models for English to Catalan (En $\rightarrow$ Ca) and English to Spanish (En $\rightarrow$ Es) translation directions. Our findings reveal pervasive gender bias across all models, with base LLMs exhibiting a higher degree of bias compared to NMT models. To combat this bias, we explore prompting engineering techniques applied to an instruction-tuned LLM. We identify a prompt structure that significantly reduces gender bias by up to 12% on the WinoMT evaluation dataset compared to more straightforward prompts. These results significantly reduce the gender bias accuracy gap between LLMs and traditional NMT systems.

Published
2024

2. Investigating the translation capabilities of Large Language Models trained on parallel data only

Author

Gilabert, Javier García, Escolano, Carlos, Savall, Aleix Sant, Fornaciari, Francesca De Luca, Mash, Audrey, Liao, Xixian, and Melero, Maite

Subjects
Computer Science - Computation and Language
Abstract

In recent years, Large Language Models (LLMs) have demonstrated exceptional proficiency across a broad spectrum of Natural Language Processing (NLP) tasks, including Machine Translation. However, previous methods predominantly relied on iterative processes such as instruction fine-tuning or continual pre-training, leaving unexplored the challenges of training LLMs solely on parallel data. In this work, we introduce PLUME (Parallel Language Model), a collection of three 2B LLMs featuring varying vocabulary sizes (32k, 128k, and 256k) trained exclusively on Catalan-centric parallel examples. These models perform comparably to previous encoder-decoder architectures on 16 supervised translation directions and 56 zero-shot ones. Utilizing this set of models, we conduct a thorough investigation into the translation capabilities of LLMs, probing their performance, the impact of the different elements of the prompt, and their cross-lingual representation space., Comment: We release our code at: https://github.com/projecte-aina/Plume

Published
2024

3. A Hard Nut to Crack: Idiom Detection with Conversational Large Language Models

Author

Fornaciari, Francesca De Luca, Altuna, Begoña, Gonzalez-Dios, Itziar, and Melero, Maite

Subjects
Computer Science - Computation and Language
Abstract

In this work, we explore idiomatic language processing with Large Language Models (LLMs). We introduce the Idiomatic language Test Suite IdioTS, a new dataset of difficult examples specifically designed by language experts to assess the capabilities of LLMs to process figurative language at sentence level. We propose a comprehensive evaluation methodology based on an idiom detection task, where LLMs are prompted with detecting an idiomatic expression in a given English sentence. We present a thorough automatic and manual evaluation of the results and an extensive error analysis.

Published
2024

4. Sequence-to-Sequence Resources for Catalan

Author

de Gibert, Ona, Kharitonova, Ksenia, Figueras, Blanca Calvo, Armengol-Estapé, Jordi, and Melero, Maite

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence
Abstract

In this work, we introduce sequence-to-sequence language resources for Catalan, a moderately under-resourced language, towards two tasks, namely: Summarization and Machine Translation (MT). We present two new abstractive summarization datasets in the domain of newswire. We also introduce a parallel Catalan-English corpus, paired with three different brand new test sets. Finally, we evaluate the data presented with competing state of the art models, and we develop baselines for these tasks using a newly created Catalan BART. We release the resulting resources of this work under open license to encourage the development of language technology in Catalan.

Published
2022

5. The Catalan Language CLUB

Author

Rodriguez-Penagos, Carlos, Armentano-Oller, Carme, Villegas, Marta, Melero, Maite, Gonzalez, Aitor, Bonet, Ona de Gibert, and Pio, Casimiro Carrino

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence, 91F20, I.2.7
Abstract

The Catalan Language Understanding Benchmark (CLUB) encompasses various datasets representative of different NLU tasks that enable accurate evaluations of language models, following the General Language Understanding Evaluation (GLUE) example. It is part of AINA and PlanTL, two public funding initiatives to empower the Catalan language in the Artificial Intelligence era., Comment: OpenCor Forum 2021. arXiv admin note: text overlap with arXiv:2107.07903

Published
2021

6. On the Multilingual Capabilities of Very Large-Scale English Language Models

Author

Armengol-Estapé, Jordi, Bonet, Ona de Gibert, and Melero, Maite

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence
Abstract

Generative Pre-trained Transformers (GPTs) have recently been scaled to unprecedented sizes in the history of machine learning. These models, solely trained on the language modeling objective, have been shown to exhibit outstanding few-shot learning capabilities in a number of different tasks. Nevertheless, aside from anecdotal experiences, little is known regarding their multilingual capabilities, given the fact that the pre-training corpus is almost entirely composed of English text. In this work, we investigate the multilingual skills of GPT-3, focusing on one language that barely appears in the pre-training corpus, Catalan, which makes the results especially meaningful; we assume that our results may be relevant for other languages as well. We find that the model shows an outstanding performance, particularly in generative tasks, with predictable limitations mostly in language understanding tasks but still with remarkable results given the zero-shot scenario. We investigate its potential and limits in extractive question-answering and natural language generation, as well as the effect of scale in terms of model size.

Published
2021

7. Are Multilingual Models the Best Choice for Moderately Under-resourced Languages? A Comprehensive Assessment for Catalan

Author

Armengol-Estapé, Jordi, Carrino, Casimiro Pio, Rodriguez-Penagos, Carlos, Bonet, Ona de Gibert, Armentano-Oller, Carme, Gonzalez-Agirre, Aitor, Melero, Maite, and Villegas, Marta

Subjects
Computer Science - Computation and Language
Abstract

Multilingual language models have been a crucial breakthrough as they considerably reduce the need of data for under-resourced languages. Nevertheless, the superiority of language-specific models has already been proven for languages having access to large amounts of data. In this work, we focus on Catalan with the aim to explore to what extent a medium-sized monolingual language model is competitive with state-of-the-art large multilingual models. For this, we: (1) build a clean, high-quality textual Catalan corpus (CaText), the largest to date (but only a fraction of the usual size of the previous work in monolingual language models), (2) train a Transformer-based language model for Catalan (BERTa), and (3) devise a thorough evaluation in a diversity of settings, comprising a complete array of downstream tasks, namely, Part of Speech Tagging, Named Entity Recognition and Classification, Text Classification, Question Answering, and Semantic Textual Similarity, with most of the corresponding datasets being created ex novo. The result is a new benchmark, the Catalan Language Understanding Benchmark (CLUB), which we publish as an open resource, together with the clean textual corpus, the language model, and the cleaning pipeline. Using state-of-the-art multilingual models and a monolingual model trained only on Wikipedia as baselines, we consistently observe the superiority of our model across tasks and settings., Comment: Accepted into Findings of ACL-IJCNLP 2021

Published
2021

8. Language Report Catalan

Author

Melero, Maite, Calvo, Blanca, Rodríguez, Mar, Villegas, Marta, Sonntag, Daniel, Editor-in-Chief, Rehm, Georg, editor, and Way, Andy, editor

Published
2023
Full Text
View/download PDF

9. Language Report Spanish

Author

Melero, Maite, Peñarrubia, Pablo, Cabestany, David, Calvo, Blanca, Rodríguez, Mar, Villegas, Marta, Sonntag, Daniel, Editor-in-Chief, Rehm, Georg, editor, and Way, Andy, editor

Published
2023
Full Text
View/download PDF

12. Factor VIII trafficking to CD4+ T cells shapes its immunogenicity and requires several types of antigen-presenting cells

Author

Kaczmarek, Radoslaw, Piñeros, Annie R., Patterson, Paige E., Bertolini, Thais B., Perrin, George Q., Sherman, Alexandra, Born, Jameson, Arisa, Sreevani, Arvin, Matthew C., Kamocka, Malgorzata M., Martinez, Michelle M., Dunn, Kenneth W., Quinn, Sean M., Morris, Johnathan J., Wilhelm, Amelia R., Kaisho, Tsuneyasu, Munoz-Melero, Maite, Biswas, Moanaro, Kaplan, Mark H., Linnemann, Amelia K., George, Lindsey A., Camire, Rodney M., and Herzog, Roland W.

Published
2023
Full Text
View/download PDF

13. Language Report Spanish

Author

Melero, Maite, primary, Peñarrubia, Pablo, additional, Cabestany, David, additional, Calvo, Blanca, additional, Rodríguez,, Mar, additional, and Villegas, Marta, additional

Published
2023
Full Text
View/download PDF

14. English-Catalan Neural Machine Translation in the Biomedical Domain through the cascade approach

Author

Costa-jussà, Marta R., Casas, Noe, and Melero, Maite

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence
Abstract

This paper describes the methodology followed to build a neural machine translation system in the biomedical domain for the English-Catalan language pair. This task can be considered a low-resourced task from the point of view of the domain and the language pair. To face this task, this paper reports experiments on a cascade pivot strategy through Spanish for the neural machine translation using the English-Spanish SCIELO and Spanish-Catalan El Peri\'odico database. To test the final performance of the system, we have created a new test data set for English-Catalan in the biomedical domain which is freely available on request., Comment: Full workshop proceedings can be found at https://multilingualbio.bsc.es/wp-content/uploads/2018/03/LREC-2018-PROCEEDINGS-MultilingualBIO.pdf

Published
2018

16. Redundancy in Innate Immune Pathways That Promote CD8 + T-Cell Responses in AAV1 Muscle Gene Transfer.

Author

Li, Ning, Kumar, Sandeep R. P., Cao, Di, Munoz-Melero, Maite, Arisa, Sreevani, Brian, Bridget A., Greenwood, Calista M., Yamada, Kentaro, Duan, Dongsheng, and Herzog, Roland W.

Subjects
DOUBLE-stranded RNA, ANTIBODY formation, GENETIC transformation, ADENO-associated virus, GENE therapy, TRANSGENE expression
Abstract

While adeno-associated viral (AAV) vectors are successfully used in a variety of in vivo gene therapy applications, they continue to be hampered by the immune system. Here, we sought to identify innate and cytokine signaling pathways that promote CD8+ T-cell responses against the transgene product upon AAV1 vector administration to murine skeletal muscle. Eliminating just one of several pathways (including DNA sensing via TLR9, IL-1 receptor signaling, and possibly endosomal sensing of double-stranded RNA) substantially reduced the CD8+ T-cell response at lower vector doses but was surprisingly ineffective at higher doses. Using genetic, antibody-mediated, and vector engineering approaches, we show that blockade of at least two innate pathways is required to achieve an effect at higher vector doses. Concurrent blockade of IL-1R1 > MyD88 and TLR9 > MyD88 > type I IFN > IFNaR pathways was often but not always synergistic and had limited utility in preventing antibody formation against the transgene product. Further, even low-frequency CD8+ T-cell responses could eliminate transgene expression, even in MyD88- or IL-1R1-deficient animals that received a low vector dose. However, we provide evidence that CpG depletion of vector genomes and including TLR9 inhibitory sequences can synergize. When this construct was combined with the use of a muscle-specific promoter, transgene expression in muscle was sustained with minimal local or systemic CD8+ T-cell response. Thus, innate immune avoidance/blockade strategies by themselves, albeit helpful, may not be sufficient to prevent destructive cellular responses in muscle gene transfer because of the redundancy of immune-activating pathways. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

19. Residual Dropout: A Simple Approach to Improve Transformer’s Data Efficiency

Author

Barcelona Supercomputing Center, Escolano, Carlos, De Luca Fornaciari, Francesca, Melero, Maite, Barcelona Supercomputing Center, Escolano, Carlos, De Luca Fornaciari, Francesca, and Melero, Maite

Abstract

Transformer models often demand a vast amount of training data to achieve the desired level of performance. However, this data requirement poses a major challenge for low-resource languages seeking access to high-quality systems, particularly in tasks like Machine Translation. To address this issue, we propose adding Dropout to Transformer’s Residual Connections. Our experimental results demonstrate that this modification effectively mitigates overfitting during training, resulting in substantial performance gains of over 4 BLEU points on a dataset consisting of merely 10 thousand examples, Peer Reviewed, Postprint (published version)

Published
2024

20. Role of FoxP3+ Regulatory T Cells in Modulating Immune Responses to Adeno-Associated Virus Gene Therapy.

Author

Muñoz-Melero, Maite and Biswas, Moanaro

Subjects
*REGULATORY T cells, *T cells, *GENE therapy, *GENETIC vectors, *ADENO-associated virus, *IMMUNE response, *CYTOTOXIC T cells
Abstract

Adeno-associated virus (AAV) gene therapy is making rapid strides owing to its wide range of therapeutic applications. However, development of serious immune responses to the capsid antigen or the therapeutic transgene product hinders its full clinical impact. Immune suppressive (IS) drug treatments have been used in various clinical trials to prevent the deleterious effects of cytotoxic T cells to the viral vector or transgene, although there is no consensus on the best treatment regimen, dosage, or schedule. Regulatory T cells (Tregs) are crucial for maintaining tolerance against self or nonself antigens. Of importance, Tregs also play an important role in dampening immune responses to AAV gene therapy, including tolerance induction to the transgene product. Approaches to harness the tolerogenic effect of Tregs include the use of selective IS drugs that expand existing Tregs, and skew activated conventional T cells into antigen-specific peripherally induced Tregs. In addition, Tregs can be expanded ex vivo and delivered as cellular therapy. Furthermore, receptor engineering can be used to increase the potency and specificity of Tregs allowing for suppression at lower doses and reducing the risk of disrupting protective immunity. Because immune-mediated toxicities to AAV vectors are a concern in the clinic, strategies that can enhance or preserve Treg function should be considered to improve both the safety and efficacy of AAV gene therapy. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

21. Viral Vector Based Immunotherapy for Peanut Allergy.

Author

Gonzalez-Visiedo, Miguel, Herzog, Roland W., Munoz-Melero, Maite, Blessinger, Sophia A., Cook-Mills, Joan M., Daniell, Henry, and Markusic, David M.

Subjects
PEANUT allergy, REGULATORY T cells, FOOD allergy, IMMUNOLOGICAL tolerance, ADENO-associated virus, TRANSGENE expression
Abstract

Food allergy (FA) is estimated to impact up to 10% of the population and is a growing health concern. FA results from a failure in the mucosal immune system to establish or maintain immunological tolerance to innocuous dietary antigens, IgE production, and the release of histamine and other mediators upon exposure to a food allergen. Of the different FAs, peanut allergy has the highest incidence of severe allergic responses, including systemic anaphylaxis. Despite the recent FDA approval of peanut oral immunotherapy and other investigational immunotherapies, a loss of protection following cessation of therapy can occur, suggesting that these therapies do not address the underlying immune response driving FA. Our lab has shown that liver-directed gene therapy with an adeno-associated virus (AAV) vector induces transgene product-specific regulatory T cells (Tregs), eradicates pre-existing pathogenic antibodies, and protects against anaphylaxis in several models, including ovalbumin induced FA. In an epicutaneous peanut allergy mouse model, the hepatic AAV co-expression of four peanut antigens Ara h1, Ara h2, Ara h3, and Ara h6 together or the single expression of Ara h3 prevented the development of a peanut allergy. Since FA patients show a reduction in Treg numbers and/or function, we believe our approach may address this unmet need. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

22. Automatic Removal of Identifying Information in Official EU Languages for Public Administrations: The MAPA Project

Author

Gianola, Lucie, primary, Ajausks, Ēriks, additional, Arranz, Victoria, additional, Bendahman, Chomicha, additional, Bié, Laurent, additional, Borg, Claudia, additional, Cerdà, Aleix, additional, Choukri, Khalid, additional, Cuadros, Montse, additional, De Gibert, Ona, additional, Degroote, Hans, additional, Edelman, Elena, additional, Etchegoyhen, Thierry, additional, Franco Torres, Ángela, additional, García Hernandez, Mercedes, additional, García Pablos, Aitor, additional, Gatt, Albert, additional, Grouin, Cyril, additional, Herranz, Manuel, additional, Kohan, Alejandro Adolfo, additional, Lavergne, Thomas, additional, Melero, Maite, additional, Paroubek, Patrick, additional, Rigault, Mickaël, additional, Rosner, Mike, additional, Rozis, Roberts, additional, Van Der Plas, Lonneke, additional, Vīksna, Rinalds, additional, and Zweigenbaum, Pierre, additional

Published
2020
Full Text
View/download PDF

23. Lifestyle Modification Improves Insulin Resistance and Carotid Intima-Media Thickness in a Metabolically Healthy Obese Prepubescent Population

Author

Gallardo-Escribano, Cristina, Vargas-Candela, Antonio, Vilches-Perez, Alberto, Muñoz-Melero, Maite, Ruiz-Moreno, M. Isabel, Benitez-Porres, Javier, Romance-Garcia, Angel Ramon, Rodriguez-Ortega, Rocío, Rosa-López, Andrés, Rosales-Jaime, Antonio, Diaz-Ruiz, Juan, Tinahones, Francisco J., Gómez-Huelgas, Ricardo, and Bernal-Lopez, M. Rosa

Published
2021
Full Text
View/download PDF

25. The Strategic impact of META-NET on the regional, national and international level

Author

Rehm, Georg, Uszkoreit, Hans, Ananiadou, Sophia, Bel, Núria, Bielevičienè, Audroné, Borin, Lars, Branco, António, Budin, Gerhard, Calzolari, Nicoletta, Daelemans, Walter, Garabík, Radovan, Grobelnik, Marko, García-Mateo, Carmen, van Genabith, Josef, Hajič, Jan, Hernáez, Inma, Judge, John, Koeva, Svetla, Krek, Simon, Krstev, Cvetana, Lindén, Krister, Magnini, Bernardo, Mariani, Joseph, McNaught, John, Melero, Maite, Monachini, Monica, Moreno, Asunción, Odijk, Jan, Ogrodniczuk, Maciej, Pęzik, Piotr, Piperidis, Stelios, Przepiórkowski, Adam, Rögnvaldsson, Eiríkur, Rosner, Mike, Pedersen, Bolette Sandford, Skadiņa, Inguna, De Smedt, Koenraad, Tadić, Marko, Thompson, Paul, Tufiş, Dan, Váradi, Tamás, Vasiļjevs, Andrejs, Vider, Kadri, and Zabarskaitè, Jolanta

Published
2016

26. TLR9-independent CD8+T cell responses in hepatic AAV gene transfer through IL-1R1-MyD88 signaling

Author

Kumar, Sandeep R.P., Biswas, Moanaro, Cao, Di, Arisa, Sreevani, Muñoz-Melero, Maite, Lam, Anh K., Piñeros, Annie R., Kapur, Reuben, Kaisho, Tsuneyasu, Kaufman, Randal J., Xiao, Weidong, Shayakhmetov, Dmitry M., Terhorst, Cox, de Jong, Ype P., and Herzog, Roland W.

Abstract

Upon viral infection of the liver, CD8+T cell responses may be triggered despite the immune suppressive properties that manifest in this organ. We sought to identify pathways that activate responses to a neoantigen expressed in hepatocytes, using adeno-associated viral (AAV) gene transfer. It was previously established that cooperation between plasmacytoid dendritic cells (pDCs), which sense AAV genomes by Toll-like receptor 9 (TLR9), and conventional DCs promotes cross-priming of capsid-specific CD8+T cells. Surprisingly, we find local initiation of a CD8+T cell response against antigen expressed in ∼20% of murine hepatocytes, independent of TLR9 or type I interferons and instead relying on IL-1 receptor 1-MyD88 signaling. Both IL-1α and IL-1β contribute to this response, which can be blunted by IL-1 blockade. Upon AAV administration, IL-1-producing pDCs infiltrate the liver and co-cluster with XCR1+DCs, CD8+T cells, and Kupffer cells. Analogous events were observed following coagulation factor VIII gene transfer in hemophilia A mice. Therefore, pDCs have alternative means of promoting anti-viral T cell responses and participate in intrahepatic immune cell networks similar to those that form in lymphoid organs. Combined TLR9 and IL-1 blockade may broadly prevent CD8+T responses against AAV capsid and transgene product.

Published
2024
Full Text
View/download PDF

27. METIS-II: low resource machine translation

Author

Carl, Michael, Melero, Maite, Badia, Toni, Vandeghinste, Vincent, Dirix, Peter, Schuurman, Ineke, Markantonatou, Stella, Sofianopoulos, Sokratis, Vassiliou, Marina, and Yannoutsou, Olga

Abstract

Abstract: METIS-II was an EU-FET MT project running from October 2004 to September 2007, which aimed at translating free text input without resorting to parallel corpora. The idea was to use “basic” linguistic tools and representations and to link them with patterns and statistics from the monolingual target-language corpus. The METIS-II project has four partners, translating from their “home” languages Greek, Dutch, German, and Spanish into English. The paper outlines the basic ideas of the project, their implementation, the resources used, and the results obtained. It also gives examples of how METIS-II has continued beyond its lifetime and the original scope of the project. On the basis of the results and experiences obtained, we believe that the approach is promising and offers the potential for development in various directions.

Published
2024
Full Text
View/download PDF

28. Distinct functions and transcriptional signatures in orally induced regulatory T cell populations.

Author

Biswas, Moanaro, So, Kaman, Bertolini, Thais B., Krishnan, Preethi, Rana, Jyoti, Muñoz-Melero, Maite, Syed, Farooq, Kumar, Sandeep R. P., Hongyu Gao, Xiaoling Xuei, Terhorst, Cox, Daniell, Henry, Sha Cao, and Herzog, Roland W.

Subjects
REGULATORY T cells, CELL populations, ORAL drug administration, CELL analysis, T cells, SCURFIN (Protein)
Abstract

Oral administration of antigen induces regulatory T cells (Treg) that can not only control local immune responses in the small intestine, but also traffic to the central immune system to deliver systemic suppression. Employing murine models of the inherited bleeding disorder hemophilia, we find that oral antigen administration induces three CD4+ Treg subsets, namely FoxP3+LAP-, FoxP3+LAP+, and FoxP3- LAP+. These T cells act in concert to suppress systemic antibody production induced by therapeutic protein administration. Whilst both FoxP3+LAP+ and FoxP3-LAP+ CD4+ T cells express membrane-bound TGF-β (latency associated peptide, LAP), phenotypic, functional, and single cell transcriptomic analyses reveal distinct characteristics in the two subsets. As judged by an increase in IL-2Rα and TCR signaling, elevated expression of co-inhibitory receptor molecules and upregulation of the TGFβ and IL-10 signaling pathways, FoxP3+LAP+ cells are an activated form of FoxP3+LAP-Treg. Whereas FoxP3-LAP+ cells express low levels of genes involved in TCR signaling or co-stimulation, engagement of the AP-1 complexmembers Jun/Fos and Atf3 is most prominent, consistent with potent IL-10 production. Single cell transcriptomic analysis further reveals that engagement of the Jun/Fos transcription factors is requisite for mediating TGFb expression. This can occur via an Il2ra dependent or independent process in FoxP3+LAP+ or FoxP3-LAP+ cells respectively. Surprisingly, both FoxP3+LAP+ and FoxP3-LAP+ cells potently suppress and induce FoxP3 expression in CD4+ conventional T cells. In this process, FoxP3-LAP+ cells may themselves convert to FoxP3+ Treg. We conclude that orally induced suppression is dependent on multiple regulatory cell types with complementary and interconnected roles. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

29. Transfer Learning with Shallow Decoders: BSC at WMT2021’s Multilingual Low-Resource Translation for Indo-European Languages Shared Task

Author

Kharitonova, Ksenia, Gibert Bonet, Ona de, Armengol Estapé, Jordi, Rodríguez i Alvarez, Mar, Melero, Maite, and Barcelona Supercomputing Center

Subjects
Translation, Machine Translation, Language Modelling, Traducció automàtica, Transfer Learning, Informàtica::Intel·ligència artificial::Llenguatge natural [Àrees temàtiques de la UPC], Machine translating, Llengües indoeuropees
Abstract

This paper describes the participation of the BSC team in the WMT2021{'}s Multilingual Low-Resource Translation for Indo-European Languages Shared Task. The system aims to solve the Subtask 2: Wikipedia cultural heritage articles, which involves translation in four Romance languages: Catalan, Italian, Occitan and Romanian. The submitted system is a multilingual semi-supervised machine translation model. It is based on a pre-trained language model, namely XLM-RoBERTa, that is later fine-tuned with parallel data obtained mostly from OPUS. Unlike other works, we only use XLM to initialize the encoder and randomly initialize a shallow decoder. The reported results are robust and perform well for all tested languages.

Published
2021

30. Transfer Learning with Shallow Decoders: BSC at WMT2021’s Multilingual Low-Resource Translation for Indo-European Languages Shared Task

Author

Barcelona Supercomputing Center, Kharitonova, Ksenia, Gibert Bonet, Ona de, Armengol Estapé, Jordi, Rodríguez i Alvarez, Mar, Melero, Maite, Barcelona Supercomputing Center, Kharitonova, Ksenia, Gibert Bonet, Ona de, Armengol Estapé, Jordi, Rodríguez i Alvarez, Mar, and Melero, Maite

Abstract

This paper describes the participation of the BSC team in the WMT2021{'}s Multilingual Low-Resource Translation for Indo-European Languages Shared Task. The system aims to solve the Subtask 2: Wikipedia cultural heritage articles, which involves translation in four Romance languages: Catalan, Italian, Occitan and Romanian. The submitted system is a multilingual semi-supervised machine translation model. It is based on a pre-trained language model, namely XLM-RoBERTa, that is later fine-tuned with parallel data obtained mostly from OPUS. Unlike other works, we only use XLM to initialize the encoder and randomly initialize a shallow decoder. The reported results are robust and perform well for all tested languages., Postprint (author's final draft)

Published
2021

31. Automatic removal of identifying information in official EU languages for public administrations : the MAPA Project

Author

Gianola, Lucie, Ajausks, Eriks, Arranz, Victoria, Bendahman, Chomicha, Bié, Laurent, Borg, Claudia, Cerdà, Aleix, Choukri, Khalid, Cuadros, Montse, de Gibert, Ona, Degroote, Hans, Edelman, Elena, Etchegoyhen, Thierry, Franco Torres, Ángela, García de Hernández, Mercedes, García-Pablos, Aitor, Gatt, Albert, Grouin, Cyril, Herranz, Manuel, Kohan, Alejandro, Lavergne, Thomas, Melero, Maite, Paroubek, Patrick, Rigault, Mickaël, Rosner, Michael, Rozis, Roberts, van der Plas, Lonneke, Viksna, Rinalds, Zweigenbaum, Pierre, and International Conference on Legal Knowledge and Information Systems

Subjects
Legal documents -- European Union countries, Multilingual communication -- European Union countries, Multilingualism -- European Union countries, Open source software
Abstract

The European MAPA (Multilingual Anonymisation for Public Administrations) project aims at developing an open-source solution for automatic de-identification of medical and legal documents. We introduce here the context, partners and aims of the project, and report on preliminary results., peer-reviewed

Published
2020

32. Automatic removal of identifying information in official EU languages for public administrations: The MAPA Project

Author

Barcelona Supercomputing Center, Gianola, Lucie, Ajausks, Ēriks, Arranz, Victoria, Gibert, Ona de, Melero, Maite, Barcelona Supercomputing Center, Gianola, Lucie, Ajausks, Ēriks, Arranz, Victoria, Gibert, Ona de, and Melero, Maite

Abstract

The European MAPA (Multilingual Anonymisation for Public Administrations) project aims at developing an open-source solution for automatic de-identification of medical and legal documents. We introduce here the context, partners and aims of the project, and report on preliminary results., Peer Reviewed, "Article signat per 30 autors/es: Lucie Gianola, Ēriks Ajausks, Victoria Arranz, Chomicha Bendahman, Laurent Bié, Claudia Borg, Aleix Cerdà, Khalid Choukri, Montse Cuadros, Ona De Gibert, Hans Degroote, Elena Edelman, Thierry Etchegoyhen, Ángela Franco Torres, Mercedes García Hernandez, Aitor García Pablos, Albert Gatt, Cyril Grouin, Manuel Herranz, Alejandro Adolfo Kohan, Thomas Lavergne, Maite Melero, Patrick Paroubek, Mickaël Rigault, Mike Rosner, Roberts Rozis, Lonneke Van Der Plas, Rinalds Vīksna, Pierre Zweigenbaum", Postprint (published version)

Published
2020

34. Lifestyle Modification Improves Insulin Resistance and Carotid Intima‐Media Thickness in a Metabolically Healthy Obese Prepubescent Population

Author

Gallardo‐Escribano, Cristina, primary, Vargas‐Candela, Antonio, additional, Vilches‐Perez, Alberto, additional, Muñoz‐Melero, Maite, additional, Ruiz‐Moreno, M. Isabel, additional, Benitez‐Porres, Javier, additional, Romance‐Garcia, Angel Ramon, additional, Rodriguez‐Ortega, Rocío, additional, Rosa‐López, Andrés, additional, Rosales‐Jaime, Antonio, additional, Diaz‐Ruiz, Juan, additional, Tinahones, Francisco J., additional, Gómez‐Huelgas, Ricardo, additional, and Bernal‐Lopez, M. Rosa, additional

Published
2020
Full Text
View/download PDF

35. Sustainable Language Data Sharing to Support Language Equality in Multilingual Europe - Why Language Data Matters:ELRC White Paper

Author

Berzins, Aivars, Choukri, Khalid, Giagkou, Maria, Lösch, Andrea, Mazo, Helene, Piperidis, Stelios, Rigault, Mickaël, Schnur, Eileen, Small, Lilli, Genabith, Josef van, Vasiljevs, Andrejs, Adamson, Andero, Anastasiou, Dimitra, Avraamides-Haratsi, Natassa, Bel, Núria, Bódi, Zoltán, Branco, António, Budin, Gerhard, Dadurkevicius, Virginijus, Smeytere, Stijn de, Dobreva, Hrístina, Domeij, Rickard, Dunne, Jane, Eide, Kristine, Foti, Claudia, Gavriilidou, Maria, Grouas, Thibault, Gruzitis, Normund, Hajic, Jan, Heinisch, Barbara, Hoste, Verónique, Jönsson, Arne, Kakoyianni-Doa, Fryni, Kirchmeier, Sabine, Koeva, Svetla, Konturová, Lucia, Kotzian, Jürgen, Krek, Simon, Kristmannsson, Gauti, Kuhmonen, Kaisamari, Lindén, Krister, Lynn, Teresa, Magone, Armands, Mazo, Hélène, Melero, Maite, Mihailescu, Laura, Montemagni, Simonetta, Õ Conaire, Micheál, Odijk, Jan, Ogrodniczuk, Maciej, Pecina, Pavel, Olsen, Jon Arild, Pedersen, Bolette Sandford, Perez, David, Repar, Andras, Terryn, Ayla Rigouts, Rögnvaldsson, Eirikur, Rosner, Mike, Routzouni, Nancy, Soria, Claudia, Soska, Alexandra, Spiteri, Donatienne, Tadic, Marko, Tiberius, Carole, Tufis, Dan, Utka, Andrius, Vale, Paolo, van den Berg, Piet, Váradi, Tamás, Vare, Kadri, Witt, Andreas, Yvon, Francois, Ziedins, Janis, Zumrik, Miroslav, Berzins, Aivars, Choukri, Khalid, Giagkou, Maria, Lösch, Andrea, Mazo, Helene, Piperidis, Stelios, Rigault, Mickaël, Schnur, Eileen, Small, Lilli, Genabith, Josef van, Vasiljevs, Andrejs, Adamson, Andero, Anastasiou, Dimitra, Avraamides-Haratsi, Natassa, Bel, Núria, Bódi, Zoltán, Branco, António, Budin, Gerhard, Dadurkevicius, Virginijus, Smeytere, Stijn de, Dobreva, Hrístina, Domeij, Rickard, Dunne, Jane, Eide, Kristine, Foti, Claudia, Gavriilidou, Maria, Grouas, Thibault, Gruzitis, Normund, Hajic, Jan, Heinisch, Barbara, Hoste, Verónique, Jönsson, Arne, Kakoyianni-Doa, Fryni, Kirchmeier, Sabine, Koeva, Svetla, Konturová, Lucia, Kotzian, Jürgen, Krek, Simon, Kristmannsson, Gauti, Kuhmonen, Kaisamari, Lindén, Krister, Lynn, Teresa, Magone, Armands, Mazo, Hélène, Melero, Maite, Mihailescu, Laura, Montemagni, Simonetta, Õ Conaire, Micheál, Odijk, Jan, Ogrodniczuk, Maciej, Pecina, Pavel, Olsen, Jon Arild, Pedersen, Bolette Sandford, Perez, David, Repar, Andras, Terryn, Ayla Rigouts, Rögnvaldsson, Eirikur, Rosner, Mike, Routzouni, Nancy, Soria, Claudia, Soska, Alexandra, Spiteri, Donatienne, Tadic, Marko, Tiberius, Carole, Tufis, Dan, Utka, Andrius, Vale, Paolo, van den Berg, Piet, Váradi, Tamás, Vare, Kadri, Witt, Andreas, Yvon, Francois, Ziedins, Janis, and Zumrik, Miroslav

Published
2019

36. English-catalan neural machine translation in the biomedical domain through the cascade approach

Author

Costa-juss��, Marta R., Casas, Noe, Melero, Maite, Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, and Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla

Subjects
FOS: Computer and information sciences, Neural Machine Translation, Computer Science - Computation and Language, Biomedical, Computer Science - Artificial Intelligence, English-Catalan, Anglès, Català, Artificial Intelligence (cs.AI), Informàtica [Àrees temàtiques de la UPC], Traducció automàtica, Enginyeria biomèdica, Computation and Language (cs.CL), Biomedical engineering, Neural networks
Abstract

This paper describes the methodology followed to build a neural machine translation system in the biomedical domain for the English-Catalan language pair. This task can be considered a low-resourced task from the point of view of the domain and the language pair. To face this task, this paper reports experiments on a cascade pivot strategy through Spanish for the neural machine translation using the English-Spanish SCIELO and Spanish-Catalan El Peri\'odico database. To test the final performance of the system, we have created a new test data set for English-Catalan in the biomedical domain which is freely available on request., Comment: Full workshop proceedings can be found at https://multilingualbio.bsc.es/wp-content/uploads/2018/03/LREC-2018-PROCEEDINGS-MultilingualBIO.pdf

Published
2018

37. ELRI. European Language Resource Infrastructure

Author

Etchegoyhen, Thierry, Anza Porras, Borja, Azpeitia, Andoni, Martínez Garcia, Eva, Vale, Paulo, Fonseca, José Luis, Lynn, Teresa, Dunne, Jane, Gaspari, Federico, Way, Andy, Arranz, Victoria, Choukri, Khalid, Popescu, Vladimir, Neiva, Pedro, Neto, Rui, Melero, Maite, Perez, David, Branco, António, Branco, Ruben, Gomes, Luís, Etchegoyhen, Thierry, Anza Porras, Borja, Azpeitia, Andoni, Martínez Garcia, Eva, Vale, Paulo, Fonseca, José Luis, Lynn, Teresa, Dunne, Jane, Gaspari, Federico, Way, Andy, Arranz, Victoria, Choukri, Khalid, Popescu, Vladimir, Neiva, Pedro, Neto, Rui, Melero, Maite, Perez, David, Branco, António, Branco, Ruben, and Gomes, Luís

Abstract

We describe the European Language Resources Infrastructure project, whose main aim is the provision of an infrastructure to help collect, prepare and share language resources that can in turn improve translation services in Europe.

Published
2018

38. English-catalan neural machine translation in the biomedical domain through the cascade approach

Author

Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Ruiz Costa-Jussà, Marta, Casas, Noe, Melero, Maite, Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Ruiz Costa-Jussà, Marta, Casas, Noe, and Melero, Maite

Abstract

This paper describes the methodology followed to build a neural machine translation system in the biomedical domain for the English-Catalan language pair. This task can be considered a low-resourced task from the point of view of the domain and the language pair. To face this task, this paper reports experiments on a cascade pivot strategy through Spanish for the neural machine translation using the English-Spanish SCIELO and Spanish-Catalan El Periódico database. To test the final performance of the system, we have created a new test data set for English-Catalan in the biomedical domain which is freely available on request., Postprint (author's final draft)

Published
2018

39. The strategic impact of META-NET on the regional, national and international level

Author

LS OZ Taal en spraaktechnologie, ILS LLI, Rehm, Georg, Uszkoreit, Hans, Ananiadou, Sophia, Bel, Núria, Bielevičienė, Audronė, Borin, Lars, Branco, António, Budin, Gerhard, Calzolari, Nicoletta, Daelemans, Walter, Garabík, Radovan, Grobelnik, Marko, García-Mateo, Carmen, van Genabith, Josef, Hajič, Jan, Hernáez, Inma, Judge, John, Koeva, Svetla, Krek, Simon, Krstev, Cvetana, Lindén, Krister, Magnini, Bernardo, Mariani, Joseph, McNaught, John, Melero, Maite, Monachini, Monica, Moreno, Asunción, Odijk, J.E.J.M., Ogrodniczuk, Maciej, Pęzik, Piotr, Piperidis, Stelios, Przepiórkowski, Adam, Rögnvaldsson, Eiríkur, Rosner, Mike, Sandford Pedersen, Bolette, Skadiņa, Inguna, De Smedt, Koenraad, Tadić, Marko, Thompson, Paul, Tufiş, Dan, Váradi, Tamás, Vasiļjevs, Andrejs, Vider, Kadri, Zabarskaitė, Jolanta, LS OZ Taal en spraaktechnologie, ILS LLI, Rehm, Georg, Uszkoreit, Hans, Ananiadou, Sophia, Bel, Núria, Bielevičienė, Audronė, Borin, Lars, Branco, António, Budin, Gerhard, Calzolari, Nicoletta, Daelemans, Walter, Garabík, Radovan, Grobelnik, Marko, García-Mateo, Carmen, van Genabith, Josef, Hajič, Jan, Hernáez, Inma, Judge, John, Koeva, Svetla, Krek, Simon, Krstev, Cvetana, Lindén, Krister, Magnini, Bernardo, Mariani, Joseph, McNaught, John, Melero, Maite, Monachini, Monica, Moreno, Asunción, Odijk, J.E.J.M., Ogrodniczuk, Maciej, Pęzik, Piotr, Piperidis, Stelios, Przepiórkowski, Adam, Rögnvaldsson, Eiríkur, Rosner, Mike, Sandford Pedersen, Bolette, Skadiņa, Inguna, De Smedt, Koenraad, Tadić, Marko, Thompson, Paul, Tufiş, Dan, Váradi, Tamás, Vasiļjevs, Andrejs, Vider, Kadri, and Zabarskaitė, Jolanta

Published
2016

40. The strategic impact of META-NET on the regional, national and international level

Author

Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Ananiadou, Sophia, Branco, Antonio, Hajic, Jan, Hernáez, Inma, Mariani, Joseph, McNaught, John, Monachini, Monica, Moreno Bilbao, M. Asunción, Odijk, Jan, Piperidis, Stelios, Rosner, Mike, Skadina, Inguna, Tadic, Marko, Thompson, Paul, Tufis, Dan, Melero, Maite, Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Ananiadou, Sophia, Branco, Antonio, Hajic, Jan, Hernáez, Inma, Mariani, Joseph, McNaught, John, Monachini, Monica, Moreno Bilbao, M. Asunción, Odijk, Jan, Piperidis, Stelios, Rosner, Mike, Skadina, Inguna, Tadic, Marko, Thompson, Paul, Tufis, Dan, and Melero, Maite

Abstract

This article provides an overview of the dissemination work carried out in META-NET from 2010 until 2015; we describe its impact on the regional, national and international level, mainly with regard to politics and the funding situation for LT topics. The article documents the initiative's work throughout Europe in order to boost progress and innovation in our field., Peer Reviewed, Postprint (author's final draft)

Published
2016

41. Selection of correction candidates for the normalization of Spanish user generated content

Author

Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Melero, Maite, Ruiz Costa-Jussà, Marta, Lambert, Patrik, Quixal, Martí, Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla, Melero, Maite, Ruiz Costa-Jussà, Marta, Lambert, Patrik, and Quixal, Martí

Abstract

We present research aiming to build tools for the normalization of User-Generated Content (UGC). We argue that processing this type of text requires the revisiting of the initial steps of Natural Language Processing, since UGC (micro-blog, blog, and, generally, Web 2.0 user-generated texts) presents a number of nonstandard communicative and linguistic characteristics – often closer to oral and colloquial language than to edited text. We present a corpus of UGC text in Spanish from three different sources: Twitter, consumer reviews, and blogs, and describe its main characteristics. We motivate the need for UGC text normalization by analyzing the problems found when processing this type of text through a conventional language processing pipeline, particularly in the tasks of lemmatization and morphosyntactic tagging. Our aim with this paper is to seize the power of already existing spell and grammar correction engines and endow them with automatic normalization capabilities in order to pave the way for the application of standard Natural Language Processing tools to typical UGC text. Particularly, we propose a strategy for automatically normalizing UGC by adding a module on top of a pre-existing spell-checker that selects the most plausible correction from an unranked list of candidates provided by the spell-checker. To build this selector module we train four language models, each one containing a different type of linguistic information in a trade-off with its generalization capabilities. Our experiments show that the models trained on truecase and lowercase word forms are more discriminative than the others at selecting the best candidate. We have also experimented with a parametrized combination of the models by both optimizing directly on the selection task and doing a linear interpolation of the models. The resulting parametrized combinations obtain results close to the best performing model but do not improve on those results, as measured on the test set. The p, Peer Reviewed, Postprint (author's final draft)

Published
2016

42. EL FUTUR DE LES LLENGÜES EN L'ERA DIGITAL: OPORTUNITATS I BRETXA LINGÜÍSTICA.

Author

Melero, Maite

Abstract

Copyright of Journal of Language & Law / Revista de Llengua i Dret is the property of Revista de Llengua i Dret and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Published
2018
Full Text
View/download PDF

43. Factor VIII trafficking to CD4+T cells shapes its immunogenicity and requires several types of antigen presenting cells

Author

Kaczmarek, Radoslaw, Piñeros, Annie R., Patterson, Paige E., Bertolini, Thais, Perrin, George Q., Sherman, Alexandra, Born, Jameson, Arisa, Sreevani, Arvin, Matthew C., Kamocka, Malgorzata M., Martinez, Michelle M., Dunn, Kenneth W., Quinn, Sean M., Morris, Johnathan J., Wilhelm, Amelia R., Kaisho, Tsuneyasu, Munoz-Melero, Maite, Biswas, Moanaro, Kaplan, Mark H., Linnemann, Amelia K., George, Lindsey A., Camire, Rodney M., and Herzog, Roland W.

Abstract

•Helper T-cell response to coagulation factor VIII critically depends on a select set of anatomically distinct antigen presenting cells•Innate immune signaling and an antigen trafficking pattern that skirts red pulp macrophages shape the immunogenicity of FVIII

Published
2023
Full Text
View/download PDF

44. Language technologies for low resource languages : sociolinguistic and multilingual insights

Author

Doğruöz, A. Seza, Sitaram, Sunayana, Melero, Maite, Sakti, Sakriani, and Soria, Claudia

Subjects
Low Resource Languages, Computational Linguistics, Technology and Engineering, TheoryofComputation_LOGICSANDMEANINGSOFPROGRAMS, LT3, Multilingualism, Languages and Literatures
Abstract

There is a growing interest in building language technologies (LTs) for low resource languages (LRLs). However, there are flaws in the planning, data collection and development phases mostly due to the assumption that LRLs are similar to High Resource Languages (HRLs) but only smaller in size. In our paper, we first provide examples of failed LTs for LRLs and provide the reasons for these failures. Second, we discuss the problematic issues with the data for LRLs. Finally, we provide recommendations for building better LTs for LRLs through insights from sociolinguistics and multilingualism. Our goal is not to solve all problems around LTs for LRLs but to raise awareness about the existing issues, provide recommendations toward possible solutions and encourage collaboration across academic disciplines for developing LTs that actually serve the needs and preferences of the LRL communities.

Published
2022

45. Investigating the quality of static anchor embeddings from transformers for under-resourced languages

Author

Singh, Pranaydeep, De Clercq, Orphée, Lefever, Els, Melero, Maite, Sakti, Sakriani, and Soria, Claudia

Subjects
LT3, Languages and Literatures
Abstract

This paper reports on experiments for cross-lingual transfer using the anchor-based approach of Schuster et al. (2019) for English and a low-resourced language, namely Hindi. For the sake of comparison, we also evaluate the approach on three very different higher-resourced languages, viz. Dutch, Russian and Chinese. Initially designed for ELMo embeddings, we analyze the approach for the more recent BERT family of transformers for a variety of tasks, both mono and cross-lingual. The results largely prove that like most other cross-lingual transfer approaches, the static anchor approach is underwhelming for the low-resource language, while performing adequately for the higher resourced ones. We attempt to provide insights into both the quality of the anchors, and the performance for low-shot cross-lingual transfer to better understand this performance gap. We make the extracted anchors and the modified train and test sets available for future research at https://github.com/pranaydeeps/Vyaapak

Published
2022

Catalog

Books, media, physical & digital resources
See catalog results