Voir la source de 2024 Workshop on Breton Language Technologies

The CNRS laboratory IKER and the U. Bretagne Ouest UBO are organizing the '''2024 Workshop on Breton Language Technologies''', which will take place at the University of Quimper (Brittany) on June 8. 


'''Contacts''' : 
: [https://iker.cnrs.fr/melanie-jouitteau/ Mélanie Jouitteau]: melanie.jouitteau at iker.cnrs.fr
: [https://zenodo.org/search?q=milan%20rezac&l=list&p=1&s=10&sort=bestmatch Milan Rezac]: milan.rezac at iker.cnrs.fr
  

The aim of this workshop is to facilitate a meeting of minds between linguists and developers of technologies for Breton and Brittonic languages. Our objective is to foster a deeper understanding of each other's achievements and to build our collective capacity in this field.
The date has been chosen to fall between the [http://celticstudents.blogspot.com/p/conference.html Celtic Student Conference] in Brest from May 30 to June 1, and the CRBC [https://nouveau.univ-brest.fr/breton-summer-school-crbc/fr/donnees-personnelles Breton Summer School] in English beginning on June 10 in Quimper.
It will be possible to follow the event on-line.
  
We are in the process of putting together a comprehensive program that includes various interventions and thematic sessions.

Confirmed speakers and attendants so far: 
: [https://nouveau.univ-brest.fr/hcti/fr/membre/liana-ermakova Liana Ermakova] (UBO), [https://loicgrobol.github.io/ Loic Grobol] (U. Paris Nanterre), [https://www.linkedin.com/in/johannes-heinecke-571a614/ Johannes Heinecke] (Orange), [https://iker.cnrs.fr/melanie-jouitteau/ Mélanie Jouitteau] (IKER, CNRS), Gweltaz Duval-Guennoc (indépendant), [https://www.linkedin.com/in/aentem/ Alan Entem] (indépendant), [https://lacito.cnrs.fr/en/directory/tanguy-solliec/ Tanguy Solliec] (LACITO, CNRS)


== Syntax, automatized translation and code-switching ==

=== Breton annotated corpora, Autogram report ===

'''Mélanie Jouitteau''' (IKER, CNRS), for the [https://autogramm.github.io/ ANR funded team Autogramm] lead by Sylvain Kahane (Modyco, CNRS, Paris) with, for Breton, Bruno Guillaume (LORIA, INRIA), Kim Gerdes (LISN!, CNRS) et Loic Grobol (Modyco, CNRS et Université Paris Nanterre), and the [[TAL]] master projects of Salomé Chandora, Katharine Jiang, Aurélien Said Housseini (2022-2023), Yingzi Liu and Yidi Huang (2023-2024).
: 30 min + 10 min questions

I present a two years project report for [[Breton treebank II]], a collective project aiming at building an annotated ''Universal Dependencies'' (UD) corpus ([[De Marneffe & al. (2021)|De Marneffe & al. 2021]], [[Nivre & al. (2020)|Nivre & al. 2020]]), based on data extracted from the ARBRES wikigrammar ([[Jouitteau (2009-)|Jouitteau 2009-]]). The work consists in extracting data from the ARBRES wikigrammar, organizing them in the [https://universaldependencies.org/docs/format.html Conll-U format], which is readable for the constitution of the richly annotated corpus, and complete this Conll format by instructing it in dependencies. Coding is operated in SUD format, with automatized translation into the UD format. The extraction is accessible [https://arboratorgrew.elizia.net/?#/projects/keuneud_breton here on github], and the enrichment on [https://arboratorgrew.elizia.net/?#/projects/keuneud_breton Arborator].

The first 2022 extraction, 'Kenstur' had obtained a small aligned corpus that has been used for the development of two separate AI trainings for Breton<->French translations ([[Grobol (2022-)|Grobol 2022-]], [[OPLB & al. (2022)|OPLB & al. 2022]]). The first feedbacks on trainings suggest that this type of high diversity corpus improves results for training on small resource sets ([[Grobol & Jouitteau (2024a)]], Entem p.c.). The 2024 extraction, 'Keneud', is somewhat bigger, organized by dialects, and includes some of the gloses annotations. A parser trained on a corrected version of [[Tyers & Ravishankar (2018)]] has pre-annotated the dependencies, with an adaptation for it to assign dominance to the [[rannig]] of each sentence in SUD. We add coding of consonantic mutations.   

* [[Grobol & Jouitteau (2024a)|Grobol, Loïc, et Mélanie Jouitteau. 2024a]]. 'ARBRES Kenstur: A Breton-French Parallel Corpus Rooted in Field Linguistics', ''Proceedings of the Fourteenth Language Resources and Evaluation Conference'', European Language Resource Association (ELRA).
* [[Tyers & Ravishankar (2018)|Tyers, Francis M. & Vinit Ravishankar. 2018]]. 'A prototype dependency treebank for Breton', ''Actes de la conférence Traitement Automatique de la Langue Naturelle'', TALN 2018, 197-204. [http://talnarchives.atala.org/TALN/TALN-2018/25.pdf texte]. [https://github.com/UniversalDependencies/UD_Breton-KEB 2023 corrected version on github].

=== Translation: state of the art and going forward ===

'''Loic Grobol''' Modyco, U. Paris Nanterre, avec Sarah Almeida Barreto U. Sorbonne Nouvelle, Mélanie Jouitteau (IKER, CNRS)
: 30 min + 10 min questions

Le premier traducteur automatique pour le breton ([[Tyers (2009)|Tyers, 2009]]) et le corpus parallèle qui l'accompagne auront 15 ans cette année. Ses performances modestes montraient déjà qu'un tel système était possible et pouvait être utile, au moins comme aide partielle à la compréhension pour les non-locuteurs. Depuis, quelques travaux proposant des améliorations ont été publiés ([[Sánchez-Cartagena & al. (2015)|Sánchez-Cartagena & al. 2015]], [[Sánchez-Cartagena & al. (2020)|2020]]), mais sans mise à disposition de logiciels ou de ressources utilisables. Pendant quinze ans, le breton n'a ainsi pas réellement bénéficié des progrès majeurs de la traduction automatique. Grobol et Jouitteau (2024) ont ensuite publié nouveau corpus parallèle extrait de la wikigrammaire ARBRES ([[Jouitteau (2009-)|Jouitteau, 2009-2024]]) et d'un traducteur automatique moderne, aux performances significativement améliorées. Les modèles aux entrainements non-documentés et aux ressources opaques sont évidemment ici hors-sujet car ils ne nourrissent pas les avancées des modèles futurs. Le breton fait également partie des langues annoncées comme qualitativement prises en charge par certains traducteurs multilingues (GPT3.5, Baidu, etc.), mais ils profitent principalement juste de la carence en matériel d’évaluation robuste pour le breton, et de rapport de force conséquent pour les imposer. En l’état, pour les développeurs qui ne volent pas leurs données aux communautés parlantes, les performances restent bien en deçà de celles de traducteurs pour des langues bien dotées, et les corpus parallèles de breton restent dispersés, mal documentés, et de qualité incertaine.

Cette présentation rend compte des travaux actuels du stage de master II de Sarah Almeida Barreto (Sorbonne nouvelle), dirigé par Loic Grobol (U. Paris Nanterre), en consultation avec Mélanie Jouitteau (IKER, CNRS). Nous présentons un inventaire complet des corpus parallèles existants, en les soumettant à une évaluation stricte pour constituer un corpus aussi complet que possible et en le soumettant à des évaluations systématiques pour nous assurer de sa qualité. Ces ressources sont mises à disposition en ligne en paquets téléchargeables, et recensées [https://entrelangues.modyco.fr/index.php/Breton#Ressources_num%C3%A9riques sur le site Entrelangues] où leurs métadonnées peuvent être discutées par les locuteurs. Nous espérons pouvoir présenter en juin le résultat d’un premier entrainement. Ce travail permettra à tou.te.s de développer des nouveaux systèmes de traduction de meilleure qualité, de concevoir des jeux de données d'évaluation qui pourront à l'avenir servir de standards, mais également d'identifier clairement les besoins en ressources pour la traduction vers et du breton afin de guider les futurs travaux de collecte de données.

* [[Grobol & Jouitteau (2024a)|Grobol, Loïc, et Mélanie Jouitteau. 2024a]]. 'ARBRES Kenstur: A Breton-French Parallel Corpus Rooted in Field Linguistics', ''Proceedings of the Fourteenth Language Resources and Evaluation Conference'', European Language Resource Association (ELRA).
* [[Jouitteau (2009-)|Jouitteau, Mélanie. 2009–2024]]. « ARBRES, wikigrammaire des dialectes du breton et centre de ressources pour son étude linguistique formelle ». 2009–2024. http://arbres.iker.cnrs.fr.
* Sánchez-Cartagena, Víctor M., Mikel L. Forcada, et Felipe Sánchez-Martínez. 2020. « A multi-source approach for Breton–French hybrid machine translation ». In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation, 61‑70. Lisboa, Portugal: European Association for Machine Translation. https://aclanthology.org/2020.eamt-1.8.
* Sánchez-Cartagena, Víctor M., Juan Antonio Pérez-Ortiz, et Felipe Sánchez-Martínez. 2015. « A Generalised Alignment Template Formalism and Its Application to the Inference of Shallow-Transfer Machine Translation Rules from Scarce Bilingual Corpora ». Computer Speech & Language, Hybrid Machine Translation: integration of linguistics and statistics, 32 (1): 46‑90. https://doi.org/10.1016/j.csl.2014.10.003.
* [[Tyers (2010)|Tyers, Francis. 2010]]. « Rule-based Breton to French machine translation ». In Proceedings of the 14th Annual Conference of the European Association for Machine Translation. Saint Raphaël, France: European Association for Machine Translation. https://aclanthology.org/2010.eamt-1.13.
* [[Tyers (2009)|Tyers, Francis M. 2009]]. « Rule-Based Augmentation of Training Data in Breton-French Statistical Machine Translation ». In Proceedings of the 13th Annual conference of the European Association for Machine Translation. European Association for Machine Translation. https://aclanthology.org/2009.eamt-1.29.

=== Multilingual Dependency Parsing for Celtic languages and its neighbouring languages ===

'''Johannes Heinecke''', Orange 
: 30 min + 10 min questions

Dependency parsing is a typical NLP task which takes plain sentences as input and generates dependency syntax trees as output. Currently, we deploy dependency parsing in a tool chain for preprocessing customer and employee comments on products and services, in order to classify thematically. POS tagging and dependency parsing is used to identify easily "who did what" and to create nominal groups as keywords (instead of simple words). In the past, handcrafted rules and lexicons where written to make a parser work. Later statistical approaches proved far more efficient, both for transition-based and graph-parsers. Recently, notably since the advent of word-embeddings (like Word2Vec) and later context aware word embeddings such as obtained from language models like BERT, graph-parsers proved to be even better.
All statistical based approaches to dependency parsing need, training data. The Universal Dependency (UD) project provides the needed data in form of 150 treebanks in over a hundred languages. Even though some treebanks are very small (as for instance the Breton treebank [[Breton KEB]]), others are rich. In case of little or no treebank data, transfer learning on similar languages can be successful, notably with the UD data: UD data has been annotated using a single set of guidelines for all languages. For instance, the set of possible part-of-speech tags, dependency relations or morpho-syntactic features are defined universally.
Most treebanks are monolingual, if expression from other languages like film titles or geographic names which can occur in the data are not counted as bi- or multilingual. In the real world, especially for speakers of Celtic languages, code switching is everywhere. We present a multilingual dependency parsing model (graph-parser) which can parse any mixture of Welsh, Irish, Scottish-Gaelic, Manx with English or French without losing much quality with respect to a monolingual model.

== Aligned audio corpora and speech to text ==


=== tba. ===

'''Tanguy Solliec''', LACITO CNRS 
: 30 min + 10 min questions


=== Anaouder, Developing ASR tools for Breton ===

'''Gweltaz Duval-Guennoc''', independent developer 
: 30 min + 10 min questions
Automatic Speech Recognition (ASR) systems can be invaluable resources for the Breton community, benefiting both learners and proficient speakers. Technologies like SMS dictation on smartphones or automatic captioning could potentially enhance exposure to the language by incorporating it into everyday handheld devices and equipping content creators with better tools.

Several notable initiatives to develop Speech-To-Text models for Breton have been made ([https://huggingface.co/alanoix/whisper-small-br Alan Entem], [https://huggingface.co/BlueRaccoon/whisper-medium-br Holly Montalvo] ''BlueRacoonTech'', [https://huggingface.co/Marxav/wav2vec2-large-xlsr-53-breton Xavier Marjou]), but we are still far from having dependable and user-friendly software for end-users. One such initiative is [[Duval-Guennoc (2022-)|Anaouder]], which focuses particularly on creating on-device solutions to prioritize user privacy and autonomy.

We have been training models using the Kaldi framework with approximately 60 hours of transcribed audio data. These models are integrated into a Python module that offers command-line tools for real-time and continuous inference from a microphone or audio files, utilizing [https://alphacephei.com/vosk Vosk] as a backend. The code and models are available under a MIT open-source license on [https://github.com/gweltou/vosk-br GitHub] and [https://pypi.org/project/anaouder/ PyPi].

In addition, we have developed a lightweight rule-based NLP [https://github.com/gweltou/ostilhou toolkit] to streamline textual and audio data processing. This toolkit includes features such as sentence segmentation, pre-tokenization, phonetization, text normalization, and inverse-normalization. Rule-based NLP tasks remain highly relevant for low-resource languages like Breton.

Lastly, we present a desktop application in its early stages of development. This application aims to simplify the creation of transcripts in Breton while still being versatile enough to assist in data alignment and verification for the development of Breton text-audio corpora.

=== tba. ===

'''Preben Vangsberg''', PhD. U. Bangor 
: 30 min + 10 min questions