4e Journée d’études
Traduction et Qualité: méthodologies en matière d’Assurance Qualité’
7 février 2014
Organisateurs : Ilse D
EPRAETEREet Carmen N
ÚÑEZ‐ L
AGOSUMR 8163 STL
Université Lille 3, Maison de la recherche, salle des colloques (F0.44)
ERREURS DE TRADUCTION, CORPUS, ET QUALITÉ / TRANSLATION ERRORS, CORPORA AND QUALITY
9.30‐10.00 : Carmen NÚÑEZ‐ LAGOS (UMR 8163 STL), Variété des connecteurs et qualité dans la traduction esp‐fr de “Cinco horas con Mario” par les apprenants.
10.00‐10.30 : Rudy LOOCK (UMR 8163 STL), Using corpora in the classroom to improve translation quality: report on an experiment
10.30‐10.45: questions ‐ question time 10.45‐11.15 : pause – break.
11.15‐11.45 : Dorothy KENNY (Dublin City University, Centre for Translation and Textual Studies), Is it getting better ? Student experiences of machine translation evaluation.
11.45 ‐12.15: Guillaume WISNIEWSKI (Université Paris Sud, Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur, LIMSI), Conception et analyse d'un corpus de post‐
éditions : estimation de qualité et analyse d'erreurs pour la traduction automatique
12.15‐12.45 : Ilse DEPRAETERE (UMR 8163 STL) & Arda TEZCAN (CrossLang) On the correlation between two kinds of quality assessment (post‐editing behaviour and human evaluation): a case study
12.45‐13.00: questions ‐ question time 13.00‐14.15 : déjeuner – lunch.
ASSURANCE QUALITÉ EN TERMINOLOGIE ‐QA AND TERMINOLOGY
14.15‐14.45 : John BEAVEN (Quality Policy Coordinator at the Translation Directorate, Council of the European Union), Quality Assurance in a large translation organisation: the experience of the Council of the EU
14.45‐15.15 : Ingrid SWINNEN (Terminology coordinator at the Translation Directorate, Council of the European Union), Best practices of terminology work in the Council of the EU.
15.15‐15.30 : questions – question time.
15.30‐16.00 : pause – break.
16.00‐16.30 : Michael WETZEL (Coreon), Guarenteeing and improving the quality of terminology resources
16.00‐16.30 : Tatiana MILLIARESSI (UMR 8163 STL), Terminologie en traduction en Sciences Humaines
16.30‐16.45 : questions – question time.
A
BSTRACTSCarmen Núñez‐ Lagos, UMR 8163 STL, Variété des connecteurs et qualité dans la traduction esp‐fr de
“Cinco horas con Mario” par les apprenants.
Abstract: Nous présenterons les résultats des traductions faites par les étudiants du Master TSM de Lille3 (Traduction Spécialisée Multilingue) d’extraits du roman espagnol Cinco horas con Mario. Nous nous focaliserons notamment sur la traduction des phrases avec le connecteur espagnol que (en tournure indépendante, ex. a), lequel n’a pas d’équivalent strict en français. Nous examinerons la variété et la fréquence des connecteurs français employés pragmatiquement équivalents (et, car, parce que, mais…), ainsi que les cas où il n’est pas rendu verbalement (omission ou signe de ponctuation). Nous tenterons de voir dans quelle mesure une variété plus ou moins importante de connecteurs français peut être en corrélation avec la qualité globale de la traduction (note finale). Nous ferons un comparatif avec la traduction proposée par Anne Robert‐Monnier (Éditions de La Découverte).
(a) Mientras sirvo el té... que los chicos tienen hambre. (En attendant, je te sers le thé…car les enfants ont faim).
Bio: http://stl.recherche.univ‐lille3.fr/sitespersonnels/nunez_lagos/index.html
Rudy Loock, UMR 8163 STL, Using corpora in the classroom to improve translation quality: report on an experiment
Abstract: The aim of this presentation is to report on an experiment with Master’s students in translation.
After translating a text at the beginning of the academic year with all the tools they found to be relevant, the students were taught on how to compile and use electronic corpora to uncover systemic inter‐
language differences in usage between two languages, English and French. The type of text retained for this study is obituaries of celebrities. Students compiled their own corpus of obituaries in original English and original French (DIY or disposable corpus) and compared lexical and grammatical usage between the two languages, with a specific emphasis on referring expressions to the deceased: First name+Family Name (e.g. Michael Jackson), Family Name (e.g. Jackson), Title+Name (e.g. Mr. Jackson), First Name (e.g.
Michael), Nickname (e.g. Bambi), Lexical Descriptions with or without Names (e.g. the American singer), Pronominal Forms (e.g. he). After observing the differences between the two languages, students were asked to revise their first translation and provide feedback on the types of changes that were made.
Selected references
Baker, M. 1993. “Corpus Linguistics and Translation Studies: Implications and Applications”. In Text and Technology, M. Baker et alii (eds), 233‐250. Amsterdam/Philadelphia: John Benjamins.
Bowker, L. 2003. “Corpus‐based applications for translator training: exploring the possibilities”. In Corpus‐
based Approaches to Contrastive Linguistics and Translation Studies, S. Granger, J. Lerot and S. Petch‐
Tyson (eds), 169‐183. Amsterdam/New York: Rodopi.
Johansson, S. and Oksefjell, S. (eds). 1998. Corpora and Cross‐Linguistic Research: Theory, Method and Case Studies. Amsterdam/New York: Rodopi.
Kübler, N., Bordet, G. and Pecman, M. 2010. “La linguistique de corpus entretient‐elle d’étroites relations avec la traduction pragmatique?”. In Passeurs de mots, passeurs d’espoir: lexicologie, terminologie et traduction face au défi de la diversité. M. Van Campenhoudt, T. Lino, and R. Costa (eds.), 579‐592.
Paris: AUF.
Laviosa, S. 2002. Corpus‐based translation studies: theory, findings, applications. Amsterdam: Rodopi.
Loock, R. 2010. “Using corpora to define target‐language use in translation”. Paper presented at the UCCTS 2010 Conference, Edge Hill University, United Kingdom, 27‐29 July 2010.
Loock, R., Mariaule, M., and Oster, C. 2013. “Traductologie de corpus et qualité: étude de cas.” Paper presented at the Tralogy 2 Conference, Paris, France, 17‐18 January 2013.
Loock, R. and Lefebvre‐Scodeller, C. To appear. “Writing about the dead: a corpus‐based study on how to refer to the deceased in French vs. English obituaries”.
Rabadán, R., Labrador, B., and Ramon, N. 2009. “Corpus‐based contrastive analysis and translation universals. A tool for translation quality assessment English‐Spanish”. Babel 55(4): 303‐328.
Salkie, R. 1997. Naturalness and contrastive linguistics. In Proceedings of PALC 97, Lodz, University of Lodz, Dept. of English, J. Melia, and B. Lewandowska‐Tomaszczyk (eds.), 297‐312. Reprinted in
Corpus Linguistics (Critical Concepts in Linguistics) Vol 4 W. Teubert and R. Krishnamurthy (eds.), 2007, 336‐351. London: Routledge.
Varantola, K. 2003. “Translators and disposable corpora”. In Corpora in Translator Education. F. Zanettin, S. Bernardini and D. Stewart (eds.), 55‐70. Manchester: St. Jerome.
Zanettin, F., Bernardini, S. and Stewart, D. (eds.). 2003. Corpora in Translator Education, Manchester/Northampton: St Jerome.
Bio : Rudy Loock is Professor of English Linguistics and Translation Studies in the Applied Languages department of the University of Lille 3 and affiliated with the CNRS laboratory Savoirs, Textes, Language (UMR 8163). His research interests include discourse pragmatics, information‐packaging, the discourse‐
prosody interface, corpus‐based translation studies, and translation quality. He is the coordinator of the CorTEx project (Corpus, Translation, Exploration) and is convinced that corpus tools can improve translation quality.
Ilse Depraetere UMR 8163 STL & Arda Tezcan, CrossLang : On the correlation between two kinds of quality assessment (post‐editing behaviour and human evaluation): a case study
Abstract: In this presentation, we will address the correlation between post‐editing similarity and the human evaluation of machine translation. On the basis of a sample of post‐edited translations and a human evaluation of the corresponding MT output, we checked whether a high similarity score corresponded to a high quality score and vice versa. A group of translation trainees post‐edited a sample and a number of the informants also rated the MT output for quality on a five‐point scale. We calculated Pearson’s correlation coefficient as well as the relative standard deviation per informant for each activity with a view to determining which of the two evaluation methods appeared to be the most reliable measurement given the project settings. The sample also enabled us to study 'post‐editing profiles' as they emerged from the post‐editing activity.
Bio: Ilse Depraetere http://perso.univ‐lille3.fr/~idepraetere
Bio: Arda Tezcan holds a master's degree in Artificial Intelligence with a specialization on Speech and Language Technology. He started his carreer as a translation engineer in the translation industry where he has gained hands‐on experience with SMT implementation in translation workflows. He has been working as a machine translation specialist at CrossLang since 2011 and has been actively involved in MT implementation and customization of various types of MT systems.
Dorothy Kenny, Dublin City University, Centre for Translation and Textual Studies, Is it getting better? Student experiences of machine translation evaluation
Abstract: Machine Translation evaluation is challenging: human evaluation of MT outputs can be expensive and subjective, while automatic evaluation metrics can be limited in their scope and difficult to interpret. In this paper I reflect on how two cohorts of students at Dublin City University (DCU) cope in their attempts to evaluate Statistical Machine Translation (SMT) outputs using both human and automatic metrics. The 60 students in question completed the MA in Translation Studies or the MSc in Translation Technology at DCU in either 2012 or 2013. As part of a module in translation technology students were required to build their own SMT engines, using cloud‐based SMT services, and to evaluate outputs from same. For the evaluation task they had at their disposal: automatic evaluation metrics as computed at system level by the SMT systems themselves; automatic evaluation metrics as computed at segment level by the Asiya‐Online Toolkit; and human evaluations conducted (also at segment level) by the students themselves. Such human evaluations were primarily based on fluency and adequacy scores, or error typologies adapted from existing sources. In this paper I report in particular on the challenges faced by students and the choices they made in evaluating outputs. Based on a qualitative analysis of students’
written assignments, I ask to what extent students embraced different methods in machine translation evaluation, and to what extent their general approach suggests emerging expertise in machine translation evaluation. Given that our students, who are on the cusp of becoming professional translators, constitute a potentially important user group for companies selling cloud‐based SMT services, such questions are important: if we can work out what translation quality metrics (and especially automatic evaluation metrics) ‘mean’ to our students, then we might be able to work out what such metrics are likely to mean to other translator‐users of cloud‐based SMT systems.
Bio: http://www.dcu.ie/info/staff_member.php?id_no=986
Guillaume Wisniewski, Université Paris Sud, Conception et analyse d'un corpus de post‐éditions : estimation de qualité et analyse d'erreurs pour la traduction automatique
Résumé: Avec le développement de la post‐édition, de plus en plus de corpus contenant des corrections de traductions sont disponibles. Ce travail présente un corpus de corrections d'erreurs de traduction collecté dans le cadre du projet ANR/TRACE et illustre les différents types d'analyses auxquels il peut servir. Nous nous intéresserons notamment à la détection des erreurs fréquentes en traduction automatique, à l'analyse de la variabilité des post‐éditions et au développement de système de prédiction de la qualité d'une traduction.
Bio: Guillaume Wisniewski est maître de conférences à l'Université Paris Sud. Ses recherches portent sur la traduction automatique et plus particulièrement sur les problèmes d'évaluation et d'estimation de confiance.
John Beaven, Quality Policy Coordinator at the Translation Directorate, Council of the European Union, Quality Assurance in a large translation organisation: the experience of the Council of the EU
Abstract: Translation in large institutions poses specific quality assurance problems, but also offers unique opportunities to address them. In this talk I will describe some of the measures put in place at the Council of the European Union, which, with the Commission and Parliament, is one of the three institutions involved in producing EU legislation.
Bio: After obtaining a PhD in Machine Translation from Edinburgh University, John Beaven worked as a computational linguistics researcher in academia (Cambridge University) and industry (Sharp Laboratories of Europe). Since 1996, he has been working in the translation departments of various EU Institutions and bodies. He is currently responsible for the Quality Policy at the Translation Directorate of the Council of the European Union.
Ingrid Swinnen, Terminology coordinator at the Translation Directorate, Council of the European Union, Best practices of terminology work in the Council of the EU.
Abstract: Translators working with texts from many different domains, like those at the Council, often spend a substantial part of their working time trying to understand the terminology used in the original document or doing research to find appropriate translations into their own languages. A well‐organised terminology service drastically reduces this extra work and allows translators to concentrate on the job of translating, and thus increases their productivity. Terminology aims to provide translators with correct solutions for their problems. This validated terminology allows texts in the target language to be as close to the original as possible, while still conforming with actual usage in the target language. The use of consistent and standardised terminology, also by drafters, increases the clarity and precision of the text.
The result is clearer texts and unambiguous communication, which are particularly important in legal matters.
Bio: Ingrid Swinnen has been working at the Council of the European Union since 1995. She started her career in the French language unit, first as translator/revisor, then as main terminologist. Since April 2009 she has been working as one of the terminology coordinators of the Translation Directorate. Her working languages include French, Dutch, English, German and Swedish.
Michael Wetzel, Coreon, Guarenteeing and improving the quality of terminology resources
Abstract: For many years we are now collecting terminology resources and storing them in database systems. Now, the terminology world sees large repositories, often covering many languages. But, with the data growth we also often recognise that the quality fails to keep up. Termbase maintainers have difficulties to control the content, are faced with inconsistencies, redundancies and varying quality. Users hesitate to rely on them when writing or translating texts, the acceptance is lost.
Recent developments make it now possible to automatically measure, improve and guarantee an enduring high quality of such resources. Further, by introducing taxonomical methods users add structure and regain control over large collections of terms. Thus the resource remains always valuable and reliable.
Bio: Michael is Managing Director of Coreon GmbH, a Berlin based language technology company. Up to 2011 he product‐managed for more than ten years SDL MultiTerm, a market leading terminology product with an install base of more than 100,000 users. His educational background is in computational linguistics.
Tatiana Milliaressi, UMR 8163 STL, Terminologie en traduction en Sciences Humaines Abstract:
Bio: http://stl.recherche.univ‐lille3.fr/sitespersonnels/milliaressi/accueilmilliaressi.html
Programme et résumés / Program and abstracts
http://stl.recherche.univ‐lille3.fr/colloques/20132014/JE_Traduction_2014_Speakers.pdf
Inscription / Registration
http://commandes.recherche.univ‐lille3.fr/inscription‐evenement/4JE‐Traduction‐et‐Qualite/
À l'attention des étudiants
Le déjeuner n'étant pas pris en charge, les étudiants souhaitant participer au déjeuner devront également régler avant le 25 janvier via le site de paiement en ligne.
Contacts
ilse.depraetere@univ‐lille3.fr carmen.nunez‐lagos@univ‐lille3.fr
Contact administratif
marie‐christine.ismaiel@univ‐lille3.fr