Transducens

Research group. Universitat d'Alacant, Spain. Artificial intelligence, language technologies, machine translation.

[Juan Antonio Pérez @japer3z]

PhD opportunity! 🚨🚨 We’re hiring a 4-year predoctoral researcher to investigate unseen language acquisition in LLMs using resources such as grammar books. Join our supportive, respectful team for this challenging project.

🔗 Details and contact info: https://t.co/Th9b4COZhN

[Transducens @transducens]

@UA_Universidad Enjoy mild winters 🌞 and summers by the sea 📷 during your 2-year research stay in
@Alicante_City, ranked the second-best city in the world for living according to expats.

[Transducens @transducens]

New postdoc 🎓 opportunity! Join the 25-year-experienced @transducens group at the Universitat d'Alacant, Spain, @UA_Universidad under an #MSCA Fellowship 📷 Low-resource language technologies, machine translation, language models, data crawling...📷 https://t.co/Y3YIR21uhX

[Transducens @transducens]

Project led by our teammate Víctor M. Sánchez-Cartagena. @UA_Universitat @Google

[Transducens @transducens]

These days, Felipe Sánchez-Martínez has been in Caifornia presenting our work on generating textual resources for Mayan languages https://t.co/qr9TyuOBuh at Google's Society-Centred AI Workshop. He’s also engaged with developers of AI for social impact across the planet.

[Prompsit @Prompsit]

Arrancamos febrero con proyecto nuevo en @Prompsit 👋 #openeurollm #multilingual #opensource 👋

[Alicante Plaza @alicanteplaza]

Prompsit, del Parque Científico de la UMH participa en un plan de modelos de lengua masivos de código abierto https://t.co/haQQrPdAeE

[Transducens @transducens]

ALIA public AI models just released in Spain partly use one of our datasets 🙂

[Juan Antonio Pérez @japer3z]

The dataset PILAR https://t.co/TNRcY7YzJf that our @transducens research group at @UA_Universidad @UA_Universitat collected and released in 2024 was included in the training data of the ALIA machine translation models https://t.co/VO5Mayp12T for Aragonese, Aranese, and Asturian, developed by the @BSC_CNS and presented last week by the Spanish Government: https://t.co/XRQ9yfyQ3T 🥳

According to their model cards, the translation models were developed by the @BSC_CNS as part of the Shared Task on Translation into Low-Resource Languages of Spain that we successfully organized last year: https://t.co/mc3hZB1Bux

Note that we have also contributed our own multilingual model, called IBRO, designed for translation from Spanish into several Romance languages, including Aragonese, Aranese, Asturian, Catalan (Valencian included), and Galician, as well as from Catalan into Aranese. The 1.3B IBRO model https://t.co/vNXXIAyIFH was publicly funded through the LiLowLa project: https://t.co/yglSh8F09J

Although the resources invested are significant, greater support from public administrations is essential. ✊🔥 Some impactful measures, like releasing certain public linguistic resources under open licenses, are not especially costly but could have a major impact.

[Juan Antonio Pérez @japer3z]

This morning, we had the pleasure of attending @AyodeleLifted's talk titled "Empowering African Languages through NLP: The Masakhane Project", where she shared the key contributions of the @MasakhaneNLP project in bringing African languages 🌍 into the digital and AI world. A truly inspiring talk that sparked plenty of interaction and engagement from the audience.

Ayodele Awokoya is currently a visiting researcher in our @transducens group, and we’re exploring the fine-tuning of large language models for machine translation involving Yorùbá, combined with data augmentation techniques.

A huge thank you also to @ELLIS_Alicante and @nuriaoliver for their amazing support and enthusiasm in hosting this event in their fantastic facilities! 🙌

This event, coincidentally, aligned with UNESCO’s World Day of African and Afrodescendant Culture, which is celebrated today. 🌟

[Transducens @transducens]

🚨 It's official! @_atoral will join our department in just a few days, thanks to the Beatriz Galindo grants, funded by the Ministry of Science, Innovation, and Universities, to attract talented researchers working abroad. We're so thrilled to start collaborating with him! 🙌🎉

[Universidad de Alicante UA @UA_Universidad]

¡Bienvenidos! 👏💪
Incorporamos nuevo talento investigador gracias a las ayudas Ramón y Cajal, Juan de la Cierva y Beatriz Galindo.
Además, sumamos cinco investigadores más con la última convocatoria del Plan GenT de la Comunitat Valenciana.

Más info en https://t.co/Y2vADF6ftm

[Universidad de Alicante UA @UA_Universidad]

¡Enhorabuena! 🙌💪
Investigadores del @DLSIUA han sido galardonados con el prestigioso Google Academic Research Award por una iniciativa que contribuirá a la preservación y promoción digital de varias lenguas mayas de Guatemala.
Más info en https://t.co/xHvNpGVHg7

[Juan Antonio Pérez @japer3z]

New day, new work to present here at #WMT24 #EMNLP24 in Miami. Today, we're presenting our neural machine translation models developed for the shared task on Translation into Low-Resource Languages of Spain. Read the full paper here: https://t.co/RTlUCMJfFi

[Juan Antonio Pérez @japer3z]

We will also be presenting today at #EMNLP24 #WMT24 the key findings of the shared task on Translation into Low-Resource Languages of Spain. A total of 17 teams from around the world contributed their models for machine translation from Spanish into Aragonese, Aranese and Asturian. Some of them will also be here presenting how they did it. Read the full paper here: https://t.co/EYxG3xmzSa

[Juan Antonio Pérez @japer3z]

Today at #WMT24 #EMNLP2024 in Miami, I'll be presenting our poster "Expanding the FLORES+ Multilingual Benchmark with Aragonese, Aranese, Asturian, and Valencian". Joint work of universities and academies: @UA_Universidad, @UOCuniversidad, @AAL_Academia, @academiaaranesa, @ALLA_ast. Aragonés, Aranés, Asturianu and Valencià take one more step into the world of artificial intelligence. Read the full paper here: https://t.co/ldtZdssqP1

[Juan Antonio Pérez @japer3z]

#EMMLP2024, here I am! 🌴 Already in Miami to present our findings from the shared task on machine translation for the Aragonese, Aranese, and Asturian languages, as well as from the process of including these languages in the FLORES+ benchmark. Excited to learn and connect in art deco style!

[Juan Antonio Pérez @japer3z]

If you hurry, you can still catch my presentation at the 13th Workshop on Natural Language Processing for Computer-Assisted Language Learning (#NLP4CALL), happening in Rennes, France, in just a few hours. I'll be presenting our paper "A conversational intelligent tutoring system for improving English proficiency of non-native speakers via debriefing of online meeting transcriptions." https://t.co/b69ZpIwyeg https://t.co/t2bucE9wxk

Our chatbot, DeMINT, uses AI to analyze online meeting transcriptions and provides personalized feedback to help non-native speakers improve their English through interactive dialogues. Early results with L1-Spanish/L2-English learners show promise for language learning in real-world settings. Code, datasets, and models on https://t.co/fEIBvBYR2Y

Project funded by @UTTERProject via cascade funding.

[Transducens @transducens]

Join the #WMT24 Shared Task on Machine Translation into Low-Resource Languages of Spain

📅 Submission Page Now Open!
https://t.co/8DcysAOTyN

About the shared task: https://t.co/PrCfGJwEHR

We look forward to your submission!

#MachineTranslation #NLP #LanguageTech

[Transducens @transducens]

We'll be presenting DeMINT's progress at the @UTTERProject User Day this Friday. Funded by UTTER via the FSTP mechanism, #DeMINT is building a chatbot that leverages LLMs to enhance your English skills via conversational analysis of what you said in online meetings.

[UTTER @UTTERProject]

🔶Join us for out 2nd UTTER User Day!

This event is fully remote and open to everyone!

Don't miss out on insights into multilingual and multimodal LLMs and NLP research.

👉 Register now: [https://t.co/DQZGRyYT92]

#UTTERProject #NLP #AI #Multilingual #VirtualMeetings

[Transducens @transducens]

Andrés Lou, representing @transducens @UA_Universidad, met with @INALIMEXICO to initiate a collaboration on the revitalization of indigenous languages in Mexico using AI-based technologies.

[INALIMEXICO @INALIMEXICO]

¡Difúndelo! 📢

El INALI y la Universidad de Alicante en España, en reunión de trabajo sobre la “Inteligencia artificial enfocada en la revitalización de las lenguas en riesgo”. Dicha presentación la realizó Andrés Lou, de dicha institución educativa.

#MéxicoPluricultural 🇲🇽

[Transducens @transducens]

Read the paper "Curated datasets and neural models for machine translation of informal registers between Mayan and Spanish vernaculars": https://t.co/CIENyKwYaB

Check the MayanV corpus for 15 Mayan languages: https://t.co/6CkJLOc48q

[Transducens @transducens]

Our researcher Andrés Lou will be presenting our contribution to the digital inclusion of Mayan languages at #NAACL2024 @naaclmeeting in Mexico City this week! Meet him at the in-person poster session on June 17th at 14:00. #mayanlanguages