Translation Colloquium
The Translation Colloquium takes place on Fridays from 10am to 11am at Kent State University/USA. Attendance is free of charge, registration is not required, all are welcome!
Recent talks on the CRITT Translation Colloquium can be found on our YouTube channel, CRITT at Kent.
Autumn 2023
Some talks will be delivered virtually, others are hybrid, i.e., in-person and streamed via the internet.
All talks are free of charge and can be attended without registration via this link
Friday, September 15, 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
The practices of translation and interpreting highlight a central aspect of human language, namely the existence of invariants, both structural and (perhaps more important) semantic, that largely motivate and constrain how we use language. Invariants play prominent formal and explanatory roles in the sciences, are heavily debated in philosophy of science, and are quietly been reconceptualized in ways that require attention and criticism by current applications of machine learning in many domains of knowledge. Using examples from recent research, I will illustrate some of the benefits and challenges of research programs in neurolinguistics that actively pursue the search for invariants and that leverage invariance to discover new phenomena, explanations, or connections between previously disjoint areas of inquiry in the language sciences.
This is a virtual talk, streamed via this link
About Giosuè Baggio
Giosuè Baggio is professor of psycholinguistics at the Norwegian University of Science and Technology in Trondheim. He is author of the books Meaning in the Brain (2018) and Neurolinguistics (2022), both published by MIT Press.
Friday, October 6, 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Julian Zapata:
“Back to Basics”: Rethinking Text Entry across World Languages in the Ubiquitous Computing and Artificial Intelligence Era
“Back to Basics”: Rethinking Text Entry across World Languages in the Ubiquitous Computing and Artificial Intelligence Era
Hybrid talk, streamed via this link and delivered in person in room SFH 112A
In this talk, we will present a research project that aims at better understanding how speakers of different languages (beyond English) produce texts today, and rethinking how they could do so in the years to come. Text entry, the process of inputting text into a computing device, is a crucial part of our interactions with technology. However, research has shown that the conventional text entry method—typing on a mechanical keyboard—is not the most efficient or appropriate way of inputting text in various contexts and for diverse users. While mobile devices can emulate a physical (mechanical) keyboard, (1) humans have not always typed, (2) not everyone is able to type, and (3) not every language is fit for typing. But how can we design more effective, accessible, and inclusive future-generation text entry methods that consider the unique characteristics of different languages, diverse users, and use cases? This fascinating question is likely to motivate writing-process and translation-process researchers alike, particularly in the age of generative artificial intelligence and neural machine translation. But we are also living in the age of speech technologies, which offer an unprecedented opportunity to “go back to basics” (as many school boards across the globe are suggesting today) in the ways we, humans, learn, write and communicate. Humans have spoken and dictated texts for millennia —why are we still typing in 2023 AC?
About Julián Zapata
Dr. Julián Zapata is an Assistant Professor of Translation Studies at the Department of Languages, Literatures and Cultures, Toronto Metropolitan University (Canada), appointed in 2022. He is also a certified translator and an entrepreneur. He completed his PhD in Translation Studies (2016) at the University of Ottawa, where he also taught for nearly a decade. Dr. Zapata has worked internationally in several projects related to translation process research, language and translation technologies, as well as translator training, with several publications and dozens of presentations, guest lectures and plenary speeches in more than 25 countries in Europe, Asia, Africa, the Middle East and the Americas. His research has been funded, notably, by the Social Sciences and Humanities Research Council of Canada, the Fonds de recherche du Québec – société et culture, and Entreprise Ireland.
Friday, October 20, 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Brian Baer:
Out from the Shadows: Tracing the Rhetoric of Indirect Translation
Out from the Shadows: Tracing the Rhetoric of Indirect Translation
This talk seeks to contribute to the growing literature on indirect translation by exploring the phenomenon in the context of twentieth and twenty-first-century Russian-English translation flows. In addition to reflecting transnational networks and the cultural (and linguistic) asymmetries embedded therein, indirect translation became central to defining "modern" translation practice, specifically, what Gideon Toury describes as the norm of directness. The first part of the paper investigates the cultural politics of relay translation in reference to two case studies: the one involving the first English translation of Dmitrii Merezhkovskii's Voskresshie bogi [Resurrected Gods], published in the early twentieth century, with Russian as the source language, and the other concerns the first English translation of Georgian dissident Levan Berzenishvili's Gulag memoir Sacred Darkness, with Russian as the intermediary language, published in the early twenty-first century. Both case studies raise the question of the relationship between relay translation and textual integrity while underscoring the persistence of the phenomenon. The second part of the paper explores the rhetoric of indirect translation in Soviet culture, as represented in theoretical works, intelligentsia discourse and works of trans-fiction, that is, fictional works featuring translators and translation. While directness claims in the pre-World War II period were used to define Soviet translation practice as superior to pre-Soviet practice, those claims were deployed in the post-War period to critique the Soviet literary bureaucracy and to expose the hypocrisy behind the official policy of "Friendship of Peoples."
Hybrid talk, streamed via this link and delivered in person in room SFH 117
About Brian Baer
Brian James Baer is Professor of Russian and Translation Studies at Kent State University. He is founding editor of the journal Translation and Interpreting Studies and co-editor of the book series Literatures, Cultures, Translation (Bloomsbury), with Michelle Woods, and Translation Studies in Translation (Routledge), with Yifan Zhu. His recent publications include the monograph Queer Theory and Translation Studies: Language, Politics, Desire and the collected volume Teaching Literature in Translation: Pedagogical Contexts and Reading Practices, with Michelle Woods. His most recent translations include Culture, Memory and History: Essays in Cultural Semiotics, by Juri Lotman, Introduction to Translation Theory, by Andrei Fedorov, Red Crosses by Sasha Filipenko, and Not Russian, by Mikhail Shevelev. He is a member of the advisory board of the Mona Baker Centre for Translation Studies, in Shanghai, China, and of the Nida Center for Advanced Research on Translation, in Rimini, Italy. He is the current president of the American Translation and Interpreting Studies Association.
Spring 2023
Some talks will be delivered virtually, others are hybrid, i.e., in-person and streamed via the internet.
All talks are free of charge and can be attended without registration
Friday, February 10, 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Guan Xingzhong:
On Harold Acton’s translation of Traditional Chinese Plays
On Harold Acton’s translation of Traditional Chinese Plays
This lecture explores Harold Acton's translation of Popular Chinese Plays. It examines the modifications Acton made in the final polishing stage of his translation manuscripts and explores possible linguistic, aesthetic and social considerations for such alterations. It is based on genetic translation criticism which refers to the reconstruction of writing process via the examination of miscellaneous information resources that bear the trace of a text's creation，such as author's notes，correspondence letters，annotations，plans，sketches，drafts，papers，manuscripts and typesets，etc. Significant in terms of epistemology and methodology for translation studies，the notion is newly introduced into the translation field to discuss translation process and translator's subjectivity.
Hybrid talk, delivered in person in room SFH 122 and streamed virtually via this link
About Guan Xingzhong
Guan Xingzhong is a professor at School of Translation and Interpreting, Beijing Language and Culture University. He is the director of Translation Research Center at the school. He was a visiting scholar of New York State University (2005-2006), Folger Shakespeare Library (2013-2014) and a short-term visiting scholar at University of Edinburgh (2019 summer). Guan’s first monograph: Speech Acts and Translation of Shakespeare Play Merry Wives of Windsor explores translation shifts involved in E-C translations. He is currently working on Translation Process Study in Harold Acton’s Manuscripts of Chinese plays. Between 2015 and 2018, he finishes the translator of Chinese version Translators through History by Judith Woodsworth and Jean Delisle. He is also responsible for the English version of Saihanba: A Champion of the Earth written by Feng Xiaojun. He has also published about 20 articles in literary translation and translation pedagogy, including An alternative explanation for translation variations in Six Chapters of a Floating Life by Lin Yutang (Perspectives: Studies in Translation Theory and Practice 2012), The pursuit of beauty by an aesthete: A study of Harold Acton's manuscripts of Popular Chinese Plays (Linguistica Antverpiensia, New Series: Themes in Translation studies, 2015) , and A Descriptive Study on the English Translation s of Ah Q Zheng Zhuan bu Lu Xun (Foreign Language Research，2020). He is also responsible for a national research project of C-E Dictionary of Chinese Traditional Plays.
Language signifies beyond its symbolic dimension. Literary production as a multisemiotic phenomenon privileges the materiality of the work--its medium--rather than only its capacity to communicate. Translators, accordingly, must see language rather than seeking to see through it. In this talk I will investigate the drama of translatorial reader response in cases of extreme translation: textual obscurity, distortion, randomness, visual noise, unfixed meaning, unstable source texts, and radical difference. We will see how the reader's--and translator's--deferred understanding are part of an aesthetic of failure in much modern art, but deferred is not defeated. In part through the prism of George Steiner's four kinds of difficulty (1978)--contingent, modal, tactical, and ontological--I consider limit cases of intransparency (asemia, glyptolalia, pataphilology), and scrutinize others (musical and pictographic languages) as attempts to aim 'beyond translation'. I will entertain cases in which the making visible of form, hypermediacy, takes precedence over meaning, and is meaning. (And perhaps hypermediacy is the new visibility.) Works employing artificial or invented languages reveal that 'unreadable' works can provide 1) a liberational poetics of defamiliarization of the everyday, 2) a vehicle of ideological ends (e.g. destroying prohibited ideas), 3) parody, 4) the sign of othering (e.g. the immigrant seeking to read the receiving culture); and 5) metacommentary on translation. Embracing translation as multisensorial, multimodal, and interrelational between form, sound, and text, and works that actively resist legibility, I engage an array of illustrative microcases from linguistic fantasies (linguistic fiction), ecoparables, visual poetry, glitch poetry, found poetry and assemblage, digital literature, text-based art, and human-aided AI literature. As María Carmen Africa Vidal Claramonte notes, "The relation between verbality and visuality in multimodal literary art, the analysis of a 'radically different linguistic order, where verbal texts are functionally taken over by visual signs' (Lee 2014), is a new and important research venue" (Translation and Contemporary Art: Transdisciplinary Encounters, 2022). Our response can be to recognize the translator's need for new multiliteracies, and see in all this a summoning of a new kind of translation, and perhaps a new kind of translator.
Hybrid talk, delivered in person in room SFH 122 and streamed virtually via this link
Dr. Washbourne is Professor of Spanish Translation at Kent State University, where he teaches translation and interpreting. His publications include the Routledge Handbook of Literary Translation (2018, co-edited), and a translation of Nobel Laureate Miguel Ángel Asturias' Legends of Guatemala (2011), for which he won the National Endowment for the Arts Translation Fellowship and the National Endowment for the Humanities Summer Stipend. He is the series editor of Translation Practices Explained (Routledge) and the new Routledge Teaching Guides. He worked for nearly two decades in migrant health interpreting and academic enhancement programs for migrant children in Ohio. He has translated Baltasar Gracián's philosophical novel, El Criticón (3 vols., 1651-1657), and is currently at work on an intellectual biography of twenty late-twentieth-century translators. His work from 2022 may be found in the Routledge Guide to Teaching Translation and Interpreting Online (introduction); The Interpreter and Translator Trainer; Aztlán: A Journal of Chicano Studies; Transletters; Translation Spaces; The Translator; TTR: traduction, terminologie, rédaction; and a co-authored study of scientific terminology in poetry (in JosTrans, The Journal of Specialised Translation) (https://www.researchgate.net/profile/Richard-Washbourne).
Friday, April, 7, 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Postponed to September, 15 2023
Since the introduction and diffusion of electrophysiology and neuroimaging in psycholinguistics about three decades ago, a large and growing body of experimental results on language in the brain have accumulated in the absence of unifying theories. Recently, machine learning methods for analyzing and modelling neural data have taken hold also in the neuroscience of language, further entrenching weak explanatory standards in our field. The response by several authoritative scholars has been to urge the formulation of theories that can effectively link computational-algorithmic descriptions of linguistic knowledge with causal-mechanistic models of language processing in the brain. In this talk, I will argue that while such multi-level theories would constitute the ultimate trophy for research in human neuroscience, for a number of methodological and epistemological reasons they do not at the moment provide a realistic path forward for the neuroscience of language. Instead, I will suggest that a more modest, progressive and effective approach can be built on three pillars: (1) a notion of plausibility to orient early theory building efforts away from hypotheses with low prior probability, (2) a search for invariants in linguistic theories that can guide inquiry into the neurobiological bases of language, and (3) a pluralistic notion of explanation, beyond causation and mechanism, that allows constructs from linguistics and neuroscience to fulfil multiple roles in the explanation process. I will illustrate this strategy with examples from recent research.
This is a virtual talk, streamed via this link
About Giosuè Baggio
Giosuè Baggio is professor of psycholinguistics at the Norwegian University of Science and Technology in Trondheim. He is author of the books Meaning in the Brain (2018) and Neurolinguistics (2022), both published by MIT Press.
Friday, April, 21 2023
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
This presentation provides an in-depth overview of modern machine interpreting, a specialized form of speech-to-speech translation that aims to produce real-time and continuous translation of spoken language.
The talk begins with a brief introduction to the different architectures used to perform this complex task, in particular, the current cascading and emerging end-to-end solutions. The technical and linguistic challenges of each approach will be highlighted, providing a high-level understanding of the workings of machine interpreting. It then delves into the results of preliminary studies on the quality of machine interpreting from a communicative perspective. This analysis will enable us to identify the potentials and limitations of current Artificial Intelligence solutions applied to the task of spoken language translation, as well as the challenges that need to be addressed in the future.
The presentation concludes with a discussion of the future perspective of this technology and its use in real-world settings.
Virtual talk, streamed via this link
Dr. Claudio Fantinuoli is a researcher and lecturer at the University of Mainz, as well as the Chief Technology Officer at KUDO. His research focuses on Natural Language Processing applied to human and machine interpreting, with a particular emphasis on computer-assisted interpreting and speech translation. He has taught Language Technologies and Conference Interpreting at various institutions in Europe and has worked as a consultant for the European Union. Dr. Fantinuoli is also the founder of InterpretBank, a computer-assisted interpreting tool for professional interpreters.
Autumn 2022
Some talks will be delivered virtually, others are hybrid, i.e., in-person and streamed via the internet.
All talks are free of charge and can be attended without registration via this link
Friday, September 9, 2022
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Neural machine translation (NMT) has been widely adopted in the translation industry in the recent years. Currently NMT use is mostly driven in a top-down fashion by language service providers (LSPs) or localization departments at large organizations, who see NMT as a method of cutting translation costs or increasing translation productivity. The aim of the OPUS-CAT project is to empower translators to utilize NMT independently in their work, so that translators can have a larger role in driving NMT adoption.
OPUS-CAT consists of an MT engine that runs locally on the translator's computer, and a collection of plugins and extensions for popular CAT tools, such as Trados, memoQ and Memsource. The local MT engine ensures the confidentiality of customer data, since no data is sent to external services. OPUS-CAT is based entirely on open-source data and software, and it offers MT models for over a thousand language pairs. These models can be fine-tuned locally with customer data. This talk will consist of two parts: 1. an introduction to OPUS-CAT and the other open-source projects that have made it possible, and 2. a discussion of the potential of NMT to change the profession of translation, and how projects like OPUS-CAT aim to direct this change in a more benign direction.
About Tommi Nieminen
Tommi Nieminen has worked as a translator since 2000, and as a translation technology developer since 2011. As a translation technology architect, he has automated translation workflows and built and deployed SMT and NMT systems for LSPs and other organizations. Currently he works as a freelance translator, developer, researcher and educator, devoting a large part of his time to the OPUS-CAT open-source project.
Virtual talk: connect to the talk via this link
Friday, September 23, 2022
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Machine translation (MT), with its dramatically improved translation performance, has become a threat not only to professional translators but also to the teachers of foreign languages. In the Japanese context of English education, it is said that the English proficiency of MTs has surpassed that of college English learners, which has called into question the significance of learning English. On the other hand, many attempts have been made to integrate MT into the foreign language teaching classroom actively. Klimova et al. (2022) provide a systematic review of pieces of literature on language teaching using MT published after 2016. The summary is as follows:
There is no doubt that MT is beneficial for foreign language learning.
MT can improve students’ L2 translation activities and experience, especially in the area of written comprehension, vocabulary in context, and as a quick language checker for short utterances.
Even students with lower proficiency can, while using NMT, develop their vocabulary and thus also enhance their writing skills.
MT allows learners to acquire vocabulary and expressions that they would not be able to come across on their own.
However, it is ideal for experienced teachers to provide a certain level of support and for students to peer review each other's work.
Students with higher proficiency can improve their metalinguistic awareness of the language differences between L1 and L2.
In this presentation, I will propose MTILT (Machine Translation in Language Teaching), which evolved from TILT (Cook, 2010). The MTILT approach draws on the elements required for professional translator training to apply to foreign language teaching. Particularly, I will propose the use of MT as a tool to support L2 writing, as sharing our preliminary analysis of the experiment conducted with college English learners in Japan.
Cook, G. (2010). Translation in language teaching. Oxford University Press.
Klimova, B., Pikhart, M., Benites, A.D. et al. (2022). Neural machine translation in foreign language teaching and learning: a systematic review. Educ Inf Technol. https://doi.org/10.1007/s10639-022-11194-2
About Masaru Yamada:
Masaru Yamada is a Professor at the College/Graduate School of Intercultural Communication at Rikkyo University. Drawing on insights from his experiences as a linguist and project manager in the translation industry, his current research explores translation processes, translation technologies (including CAT, MT, and PE), and TILT (Translation in Language Teaching). He serves as a board member of the Asia-Pacific Association for Machine Translation (AAMT), and is also a member of the Editorial Board of The Interpreter and Translator Trainer. He is a co-editor of Metalanguages for Dissecting Translation Processes: Theoretical Development and Practical Applications (Routledge, 2022).
Hybrid talk, delivered in person in room SFH 121 and streamed via: virtual talk via this link
The talk will be followed by a small reception where students can directly interact with the speaker.
Friday, October 7, 2022
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
The Covid-19 pandemic has changed language-related practices in many ways. In the world of professional interpreting, it has greatly accelerated the evolution of a practice that had hitherto been much debated yet remained marginal, namely the practice of delivering interpreting services via audio or video link, known as remote or distance interpreting. Although this modality of interpreting has obvious benefits, research has shown that it entails many challenges for interpreters, including increased fatigue, sometimes marked by a decline in output quality, a lack of rapport with those who require the interpretation, and/or a deterioration of the communicative dynamics. Whilst the full impact of remote interpreting on interpreters and users of interpreting services has yet to be determined, the pandemic has also created new and untested configurations of remote interpreting, ranging from working fully online for prolonged periods of time to using cloud-based/software-based interpreting platforms, combining multiple modes (audio-only/video) and/or devices to connect to clients and fellow interpreters, and integrating additional language technology to assist interpreters. In this presentation I will review previous research on remote interpreting in light of its relevance for examining and understanding the practice of remote interpreting as it has evolved since the beginning of the pandemic and identify those aspects that require further or renewed attention in research.
Virtual talk: connect to the talk via this link
About Sabine Braun
Dr Sabine Braun is Professor of Translation Studies and Director of the Centre for Translation Studies at the University of Surrey in the UK. She is also currently serving as Co-Director of the Surrey Institute for People-Centred Artificial Intelligence. Her research explores the integration and interaction of human and machine in translation and interpreting, for example to improve access to critical information, media content and vital public services such as healthcare and justice for linguistic-minority populations and other groups/people in need of communication support. Her overarching interest lies in the notions of fairness, trust, transparency, and quality in relation to technology use in these contexts.
Friday, October 21, 2022
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Maarit Koponen:
What is post-editing, anyway? The evolving processes of translators and machine translation
What is post-editing, anyway? The evolving processes of translators and machine translation
Machine translation plays an ever growing role in the work of translators as well as other professionals in the language industry. Traditionally, the use of machine translation by translators has been referred to as post-editing to mean a process where the translator uses the machine-generated translation as a raw version, correcting and editing as necessary. However, as technologies and workflows in the field are evolving, translators’ interaction with technology is also taking varied shapes. The technologized translation environments commonly mix different resources: machine translation, translation memories and terminology resources. Machine translation systems with interactive and adaptive features also add to the complexity of the situation. These evolving scenarios lead us to ask: What is post-editing, anyway? Where does it intersect with or differ from translating and revising? Do the established notions of how to post-edit still apply? In this talk, I examine the concept of post-editing and discuss how it may be changing in the technological and practical contexts of the language industry.
Virtual talk: connect to the talk via this link
About Maarit Koponen
Dr Maarit Koponen currently works as Professor of Translation Studies at the University of Eastern Finland. She has previously worked as a post-doctoral researcher at the University of Helsinki and as a lecturer at the University of Turku after receiving her PhD in Language Technology at the University of Helsinki in 2016. Her research focuses on the use of machine translation and other translation technologies, machine translation post-editing and quality evaluation. She has also worked as a professional translator for several years. As part of the EU COST Action "Language in the Human-Machine Era" launched in November 2020, she chairs Working Group 7 which focuses on the future of language work and language professionals.
Friday, November 18, 2022
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
The language industry is currently experiencing a tremendous sea change, driven by rapid technological advancement, increasingly ubiquitous integration of artificial intelligence in various phases of project workflows, and a noticeable proliferation in language industry career paths. This is having a direct impact on the very kinds of tasks language industry professionals undertake, the technologies they use to do so, and the stakeholders with whom they interact and collaborate in the process. In an industry where tasks are often ill-defined and ever-changing as it is, diversification of services offered both within and across language industry strands has become commonplace, to the point of almost being an expectation from a competence and skillset perspective.
The largely inherent need for diversification has sparked renewed interest among Translation Studies scholars (Risku and Schlager, 2021; Alves and da Silva, 2021) in exploring what expertise implies when it comes to language industry training and how some of the central tenets of expertise can give shape to competence frameworks. Classic conceptualizations of expertise from cognitive psychology (Ericsson and Charness, 1997) have been complemented by socio-cognitive paradigms (Muñoz, 2014) that explore expertise based on interaction with artefacts and other actants (Risku and Schlager, 2021). Discussions of domain-specific, routinized expertise (Shreve, 2002) are now accompanied by discussions of adaptive expertise (Angelone, 2022) and, more recently, interactional/team expertise. This presentation will leverage alumni survey data collected from recent graduates of Kent State University’s MA in Translation program in shedding light on current language industry realities as they pertain to adaptive expertise, ill-defined tasks, and the need for diversification. Their voices and experiences will be used a platform for discussing how we might best go about weaving adaptive expertise into competence frameworks and training activities to better prepare students for sustainable employment in the language industry.
Hybrid talk, delivered in person in room SFH 121 and streamed via: virtual talk via this link
The talk will be followed by a small reception where students can directly interact with the speaker.
About Erik Angelone
Erik Angelone is currently Professor of Translation Studies at Kent State University after having previously worked as Professor of Translator Studies at the Zurich University of Applied Sciences and Lecturer of Translation at the University of Heidelberg. He received his PhD in Translation Studies from the University of Heidelberg and his MA in Intercultural Communication from the University of Maryland Baltimore County. His research interests include language industry studies, process-oriented translation training and assessment, and cognitive processes in translation. He’s a member of the TREC thematic network on empirical research in translation and cognition.
Spring 2022
Some talks will be delivered virtually, others are hybrid, i.e., in-person and streamed via the internet.
All talks are free of charge and can be attended without registration via: https://us.bbcollab.com/guest/d5d18772714e4131b7fa2e699570f233
Friday, January 28, 2022,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Choosing the right customizable NMT for practical purposes is hard. With 40+ MT providers on the market, stock model performance varies a lot depending on the language pair and content type. Customization adds its own layer of complexity, as its success depends on the data volume, quality, and the approach to customization, which also varies across different MT platforms. As if those factors weren't already enough, different MT usage scenarios impose their own requirements, such as the necessity to provide translations with a proper tone of voice or gender.
In this talk, we present an approach used at Intento to build multi-vendor MT solutions for some of the largest global companies. We will walk you through how we address all the issues listed above while keeping it practically feasible considering limited resources and time. We will touch on data cleaning, model training, and training analysis, comparing multiple models for post-editing and real-time use-cases, and ROI estimation.
About Konstantin Savenkov:
After receiving a PhD in 2008, Konstantin Savenkov led research and development efforts for online content services, then worked as CTO at Zvooq and as a chief operating officer at Bookmate. In 2016, he contributed his experience in artificial intelligence (AI), technology, and operations to found Intento, Inc., where they build tools to source, evaluate, and use machine translation and other cognitive AI services.
Talk delivered virtually
Autumn 2021
Some talks will be delivered virtually, others are hybrid, i.e., in-person and streamed via the internet.
All talks are free of charge and can be attended without registration via: https://us.bbcollab.com/guest/d5d18772714e4131b7fa2e699570f233
Friday, December 10th, 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
The advent of Neural Machine Translation (NMT) has led to striking improvements in the quality of MT. Unfortunately, it has also produced models that are much less interpretable. Using an NMT model implies that there is no learned bilingual lexicon nor sets of translation rules that can be inspected once the model has been trained. This makes it notoriously difficult to understand why a neural model made a given error or opted for an infelicitous translation, let alone to 'fix' it.
After explaining how NMT roughly works, I will discuss how neural model 'interpretability' has become a prolific field of research on its own. I will then present some of my work in this area, showing that NMT learns to encode grammatical features of the source sentence in ways that do not always match our linguistic intuition. Finally, I will present the InDeep project, which aims to make the latest interpretability techniques accessible to users of NMT-assisted translation tools.
About Arianna Bisazza:
Arianna Bisazza is assistant professor in computational linguistics at the University of Groningen, The Netherlands. Her research aims to identify the intrinsic limitations of current language modeling paradigms, and to improve the quality of machine translation for challenging language pairs. She obtained her PhD in statistical machine translation from FBK/University of Trento in 2013, and has been recently awarded a large national research grant on 'Interpreting Deep Learning Models for Text and Sound' as part of the InDeep consortium (https://interpretingdl.github.io).
Talk delivered virtually
Friday, November, 5th 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
The construct of cognitive load, associated with the constructs of attentional resources and working memory, is central in explaining interpreting performance, including failures and interpreting tactics, inter alia through several hypotheses about challenges in the interpreting process. Its use as an ordinal variable is uncontroversial, and the effects of increasing cognitive load can be seen in the translation and interpreting product through various indicators, but using cognitive load as an interval variable is difficult. Most quantitative methods used to measure it actually measure cognitive effort. This is problematic because the relations between cognitive load and cognitive effort cannot be assumed to be straightforward, especially in the lower and higher ends of relevant cognitive effort. Triangulation with several indicators, coupled with retrospective reports, might be a promising avenue to take the investigations further. In some cases, retrospective verbal indicators could be more useful than physiological effort measurements to investigate the effects of cognitive load.
About Daniel Gile
Daniel Gile first studied mathematics (and sociology as a minor for his first degree). He later migrated to technical translation and conference interpreting and started studying Japanese. He rapidly became interested in research and earned a PhD in Japanese, then a PhD in linguistics and then a post-doctoral degree (habilitation), all related to translation and/or interpreting. He has lectured and published extensively and is a member of many editorial and consultative boards, a former CERA professor and current CERA staff member, a co-founder of the European Society for Translation Studies and its third president, the founder of the CIRIN network and the editor of the CIRIN Bulletin. His research interests include translator and interpreter training, interpreting cognition, and the training of researchers in Translation and Interpreting Studies. For more details, see the CIRIN website at https://cirin-gile.fr.
Talk delivered virtually
Friday, October 22, 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time) , virtual presentation
Gualtiero Piccinini:
Neural Computation and 4EA Cognition: Embodied, Embedded, Enactive, Extended, and Affective Neurocognitive Mechanisms
Neural Computation and 4EA Cognition: Embodied, Embedded, Enactive, Extended, and Affective Neurocognitive Mechanisms
4EA (embodied, embedded, enactive, extended, and affective) approaches to cognition are often pitched as alternatives to computational or representational approaches. I will argue that this is a false contrast. I will introduce a neurocomputational account of cognition that relies on neural representations. I will then argue that this account is not only compatible with (non-question-begging) 4EA approaches, but that it requires embodiment, embeddedness, and affectivity at its very core. Specifically, the construction of neural representations and their nonderived semantic content requires a tight interplay between nervous system, body, and environment, mediated by the reward system. I will use examples from translation studies and bilingualism.
About Gualtiero Piccinini:
Gualtiero Piccinini is Curators' Distinguished Professor of Philosophy and Associate Director of the Center for Neurodynamics at the University of Missouri - St. Louis. In 2014, he received the Herbert A. Simon Award from the International Association for Computing and Philosophy. In 2018, he received the K. Jon Barwise Prize from the American Philosophical Association. In 2019, he received the Chancellor's Award for Research and Creativity from University of Missouri - St. Louis. His publications include Physical Computation: A Mechanistic Account (Oxford University Press, 2015) and Neurocognitive Mechanisms: Explaining Biological Cognition (Oxford University Press, 2020).
Virtual talk
Friday, September 24th, 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Ellen Elias-Bursac:
Working in a Tug-of-War, Translating Evidence and Interpreting Testimony at a War Crimes Tribunal
Working in a Tug-of-War, Translating Evidence and Interpreting Testimony at a War Crimes Tribunal
This is in-person talk, scheduled for: SFH room 112/112a. It can also be attended virtually.
How can defendants be fairly tried if they cannot understand the charges being raised against them? Can a witness testify if the judges and attorneys cannot understand what the witness is saying? Can a judge decide whether to convict or acquit if she or he cannot read the documentary evidence tendered by the attorneys? The very viability of international criminal prosecution and adjudication hinges on vast amounts of translation and interpreting. At the International Criminal Tribunal for the Former Yugoslavia, witnesses testified through an interpreter about translations, attorneys argued through an interpreter about translated evidence and courtroom interpreting, and the judges adjudicated on the interpreted testimony and translated evidence. The talk will focus on what it was like for the translators and interpreters to work in this setting and how the ever-present translating and interpreting shaped the daily work of the tribunal, the trials and the judgments.
About Ellen Elias-Bursac
Ellen Elias-Bursac is a literary translator and translation studies scholar. She is also currently president of the American Literary Translators Association. She has been translating from the Bosnian, Croatian, and Serbian since the 1980s and has, to date, translated some thirty book-length works of both fiction and non-fiction, by authors such as David Albahari, Ivana Bodrožić, Daša Drndić, Kristian Novak, Igor Štiks, Dubravka Ugrešić, Karim Zaimović. The wars in the Balkans have shaped her professional life, both in terms of the writing she has translated and of her six years of experience as a translator and reviser in the language services at the International Criminal Tribunal for the former Yugoslavia (the ICTY) in The Hague. Her work as a scholar of translation studies ranges from her analysis of the role of translation at the ICTY to archival research into the backstories of the cold-war translation of books by authors from Yugoslavia.
This will be a hybrid event. It can be attended in-person and can also be attended online for free.
Ellen Elias-Bursac will also be meeting with students to talk about the American Literary Translation Association (ATLA) from 2pm - 3:30pm in room SFH 110.
Spring 2021
Due to the pandemic, these talks are delivered virtually -- but it has the advantage that we can more easily invite speakers from abroad.
All talks are free of charge and can be attended without registration via: https://us.bbcollab.com/guest/d5d18772714e4131b7fa2e699570f233
Friday, April 16th, 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
Neural MT can facilitate communication in a way that surpasses previous MT paradigms, but there are also consequences of its use. As with the development of any technology, MT is not ethically neutral, but rather reflects the values of those behind its development. This talk considers the ethical issues around MT, beginning with data gathering and reuse and looking at how MT fits with the values and codes of the translator. If machines and systems reflect value systems, can they be explicitly ‘good’ and remove bias from their output? What is the contribution of MT to discussions of sustainability and diversity? Rather than promoting an approach that involves following a set of instructions to implement a technology unthinkingly, this talk will highlight the importance of a conscious decision-making process when designing a data-driven MT workflow.
About Joss Moorkens
Joss Moorkens is Chair of postgraduate translation programmes at the School of Applied Language and Intercultural Studies at Dublin City University and a Funded Investigator at the ADAPT Centre. He has authored over 50 articles, book chapters, and conference papers on translation technology, user evaluation of machine translation, translator precarity, and translation ethics. He is General Co-Editor of the journal Translation Spaces and co-edited the book Translation Quality Assessment: From Principles to Practice (2018), and special issues of Machine Translation (2019) and Translation Spaces (2020). He sits on the board of the European Masters in Translation network and is a member of the Centre for Translation and Textual Studies.
Friday, March 19th, 2021,
10:00 a.m. –11:00 a.m., EDT (Eastern Daylight Time)
In this talk, I will present the background to a research project that aims to contribute to the understanding of how translation expertise is seen and constructed in translation practice. The research project is informed by translation workplace studies and approaches the topic primarily from a qualitative, ethnographic perspective.
The question of what translation expertise means in authentic work settings has so far been under-researched. Accordingly, the objective of this research project is to explore how expertise is socially constructed and rationalized in the field and how it manifests itself in day-to-day activities. The main focus thereby lies on the social and interactional construction of what is required to perform translators’ tasks. This includes normative descriptions as well as the actual workplace praxis observed and the interpretations or rationalizations of the various social actors involved. In the talk, I will discuss how the notion of translation expertise has been conceptualized and investigated in translation studies and describe what it means to contextualize translation expertise in the workplace.
About Hanna Risku
Hanna Risku is professor for translation studies at the Centre for Translation Studies at the University of Vienna, Austria. Her previous positions include professor for translation studies at the University of Graz, professor for applied cognitive science and technical communication at the Danube University Krems, guest professor at the University of Aarhus, Denmark, and lecturer at the University of Skövde, Sweden. Her research areas include cognitive scientific foundations of translation, cognitive ethnography, and translation workplace and network research.
Friday, February 26th, 2021,
10:00 a.m. –11:00 a.m., EST (Eastern Standard Time)
Janice Jun PAN:
From corpus-based interpreting studies to interpreting data “mining”: breaking the boundaries
From corpus-based interpreting studies to interpreting data “mining”: breaking the boundaries
Interpreting has existed since ancient times and evolved into a professional activity since the beginning of the past century. There is a large amount of interpreting-related data that is worthy of scientific exploration. Nevertheless, the systematic collection and analysis of such data were not possible until the recent advancement of corpus-based interpreting studies.
In this talk, I will start with a review of corpus-based interpreting studies, and then move on to introduce the progress of several corpora projects that I have engaged in. The first is the Chinese/English Political Interpreting Corpus (CEPIC), a 6.5-million-word-token corpus that includes transcripts of politician talks, and their translation and interpreting over a 21-year span. Two other on-going projects will also be discussed: one on student interpreters and translators, and the other on the perceived role of interpreters.
Through these examples, I hope to illustrate some possible ways to build synergies with digital humanities and data science, which will help to further advance the subdiscipline of corpus-based interpreting studies and serve a fruitful future direction.
About Janice Jun PAN
Janice Jun PAN is Associate Professor in the Department of Translation, Interpreting and Intercultural Studies at Hong Kong Baptist University. She works as Managing Editor of Bandung: Journal of the Global South and Review Editor of The Interpreter and Translator Trainer. Her research interests include digital humanities and interpreting/translation studies, corpus-based interpreting/translation studies, interpreting/translation and political discourse, learner factors in interpreter training, professionalism in interpreting, and bibliometric studies. Her articles are included in journals including Target: International Journal of Translation Studies, The Interpreter and Translator Trainer, Perspectives: Studies in Translatology, inTRAlinea, etc. She also published with Routledge, Springer, Peter Lang, etc. She recently accomplished a digital scholarship project, The Chinese/English Political Interpreting Corpus (CEPIC)
Friday, January 29th, 2021,
10:00 a.m. –11:00 a.m., EST (Eastern Standard Time)
Video games combine most of the characteristics of previous media into an interactive artefact that harnesses all the capabilities of the computer age. In this sense, traditional creativity difficulties intertwine with the interactivity that attracts players to this most immersive media. Semiotics and pragmatics offer robust analytical tools to study the layering of meaning-making assets in entertainment products and explain why the concept of 'playability' is essential to determine quality in video game localisation as expected in the industry.
About Miguel Bernal-Merino
Miguel Á. Bernal-Merino, PhD in the localisation of multimedia interactive entertainment software at Imperial College, is a localisation researcher and writer who lectures at the University of Roehampton, UK and several other universities across Europe. He is the co-founder and elected chair of the 'IGDA Localization SIG'. He has coordinated the 'Game Localization Round Table' (for Localization World), as well as the 'Localization Summit' (for GDC) from their conception.
Autumn 2020
invited talks (Program, Abstracts):
Gregory Shreve: The Strange Attractions of Translation: Translation Performance as Complex-Adaptive System
Fabio Alves: Grounding Cognitive Translation Studies within a situated, distributed and extended approach
Masaru Yamada: Post-editing and a Sustainable Future for Translator
Mirko Plitt: Towards the transparent translation supply chain?
Spence Green: Practical Experience with Interactive Machine Translation
Alon Lavie: COMET - a Neural Framework for State-of-the-Art MT Evaluation Models
Natasha Tokowicz: The Translation Ambiguity Disadvantage in Language Processing: The Influence of Context and Proficiency
Antonio Toral: Human Parity and Translation Universals in Machine Translation
Maeve Olohan: Translation Practice, Technology, and Power
Andy Way: The Effect of NMT on Translators and the Translation Process
Poster presentations: links to the Program, and to the Abstracts
Spring 2020
Canceled due to COVID-19
Friday, April 3rd, 2020,
Janice Jun PAN:
From corpus-based interpreting studies to interpreting data “mining”: breaking the boundaries
From corpus-based interpreting studies to interpreting data “mining”: breaking the boundaries
10:00 a.m. –11:00 a.m., in Satterfield Hall, room 112/112a (SFH 112)
Interpreting has existed since ancient times and evolved into a professional activity since the beginning of the past century. There is a large amount of interpreting-related data that is worthy of scientific exploration. Nevertheless, the systematic collection and analysis of such data were not possible until the recent advancement of corpus-based interpreting studies.
In this talk, I will start with a review of corpus-based interpreting studies, and then move on to introduce the progress of several corpora projects that I have engaged in. The first is the Chinese/English Political Interpreting Corpus (CEPIC), a 6.5-million-word-token corpus that includes transcripts of politician talks, and their translation and interpreting over a 21-year span. Two other on-going projects will also be discussed: one on student interpreters and translators, and the other on the perceived role of interpreters.
Through these examples, I hope to illustrate some possible ways to build synergies with digital humanities and data science, which will help to further advance the subdiscipline of corpus-based interpreting studies and serve a fruitful future direction.
About Janice Jun PAN
Janice Jun PAN is Associate Professor in the Department of Translation, Interpreting and Intercultural Studies at Hong Kong Baptist University and now Visiting Faculty at the Translation Research and Instruction Program, State University of New York at Binghamton. She works as Managing Editor of Bandung: Journal of the Global South and Review Editor of The Interpreter and Translator Trainer. Her research interests include digital humanities and interpreting/translation studies, corpus-based interpreting/translation studies, interpreting/translation and political discourse, learner factors in interpreter training, professionalism in interpreting, and bibliometric studies. Her articles are included in journals including Target: International Journal of Translation Studies, The Interpreter and Translator Trainer, Perspectives: Studies in Translatology, inTRAlinea, etc. She also published with Routledge, Springer, Peter Lang, etc. She recently accomplished a digital scholarship project, The Chinese/English Political Interpreting Corpus (CEPIC)
Friday, February 21th, 2020,
10:00 a.m. –11:00 a.m., in Satterfield Hall, room 112/112a (SFH 112)
Translation tools are becoming increasingly embedded in the translation profession. In particular, machine translation technology continues to advance, and the recent paradigm shift which saw the incorporation of artificial intelligence-based machine learning techniques – an approach referred to as Neural Machine Translation (NMT) – has been accompanied by a growing interest in incorporating this technology into the workflow of professional translators. In this presentation, we will explore the notion of “fit-for-purpose” translation, as well as the emerging concept of “machine translation literacy” – the idea that users of machine translation need to develop a critical approach to this technology in order to decide whether, when, why and how to use it in an informed way. We will consider how this can be integrated into translator training, but also how translators can exercise professional and social responsibility by helping to educate users outside the language professions to approach machine translation with a critical eye.
About Lynne Bowker
Lynne Bowker is Full Professor at the School of Translation and Interpretation at the University of Ottawa in Canada, where she also holds a cross-appointment to the School of Information Studies. Her areas of teaching and research interest include computer-aided and machine translation, corpus linguistics, terminology and language for special purposes. Her publications include Computer-Aided Translation Technology (University of Ottawa Press, 2002), Working with Specialized Language (Routledge, 2002) and Machine Translation and Global Research (Emerald, 2019)
Friday, January 31th, 2020,
the talk will take place 10:00 a.m. –11:00 a.m., in Business Administration Bldg, room 210 (BSA 210 )
A persistent criticism of radical embodied cognitive science is that it will be impossible to explain "real cognition" without invoking mental representations. This paper provides an account of explicit, real-time thinking of the kind we engage in when we imagine counter-factual situations, remember the past, and plan for the future. We first present a very general non-representational account of explicit thinking, based on pragmatist philosophy of science. We then present a more detailed instantiation of this general account drawing on nonlinear dynamics and ecological psychology. Finally, I will make some preliminary speculations about the relationship between our pragmatist views and translation process research.
(This work is a collaboration with Gui Sanches and Vicente Raja.).
About Anthony Chemero:
Anthony Chemero is Distinguished Research Professor of Philosophy and Psychology at the University of Cincinnati. His research is both philosophical and empirical; typically, it tries to be both at the same time. His research is focused on questions related to nonlinear dynamical modeling, ecological psychology, complex systems, phenomenology, and artificial life. He is the author of more than 100 articles and the books Radical Embodied Cognitive Science (2009, MIT Press) and, with Stephan Käufer, Phenomenology (2015, Polity Press). He is currently editing the second edition of The MIT Encyclopedia of the Cognitive Sciences
Autumn 2019
Friday, December 6th, 2019,
the talk will take place 10:00 a.m. –12:00 a.m., in the Business Administration Bldg, room 209 (BSA 209)
The history of translation reaches out back to thousands of years ago, and we are just witnessing possibly the first breakthrough in how we think about translation. What does it feel like to translate content in our age? What does the market demand? With the interesting developments in computer science, the big question in everyone’s mind: what is happening to the academia and profession of translation? We will go over a general overview of the past, today and future, with a conversation about which positions we can take as industry professionals.
The talk (10:00 - 11:00) is followed by a tutorial in which Oytun Tez gives some background information on the data, data analysis, and some techniques in software engineering that are in place at MotaWord.
About Oytun Tez:
Oytun is the co-founder and CTO of MotaWord, the world’s fastest business translation platform. Majored in linguistics, he is a software engineer by vocation. He grew an interest in collaborative workflows which MotaWord implements fully, and the automation of human collaboration. His most recent toys are GPT-2 algorithms, inline skating and kites.
November 1st, 2019
the talk will take place 10:00 a.m. - 11:00 a.m. in the Business Administration Bldg, room 209 (BSA 209)
Many words in a given language can be translated in more than one way into another language. This mis-mapping of translations across languages, known as “translation ambiguity”, causes difficulty in second language processing. Such difficulty has been reported for learners at the beginning stages of acquisition (e.g., Degani et al., 2014) as well as speakers of intermediate proficiency (e.g., Eddington & Tokowicz, 2013; Tokowicz & Kroll, 2007), and even balanced bilinguals (e.g., Boada et al., 2013). Translation ambiguity has also been demonstrated in a variety of cross-language pairs (e.g., Prior et al., 2007). In this presentation, I will discuss how we measure translation ambiguity, as well as differences in the types of translation ambiguity across languages, focusing especially on the difference between Mandarin Chinese, Dutch, German, and English.
About Natasha Tokowicz:
Natasha Tokowicz was born in northeastern Massachusetts and received a BA in Psychology with a minor in Spanish from the University of Massachusetts at Amherst. She earned an M.S. and Ph.D. in Cognitive Psychology from The Pennsylvania State University. She completed post-doctoral fellowships at Carnegie Mellon University and the University of Pittsburgh prior to beginning a faculty appointment at the University of Pittsburgh in 2004. She is currently Associate Professor of Psychology and Linguistics and Research Scientist at the Learning Research and Development Center at the University of Pittsburgh, with an appointment in the Center for the Neural Basis of Cognition. Her research combines behavioral and cognitive neuroscientific methodologies to address questions about adult second language learning and bilingualism. Her book, Lexical Processing and Second Language Acquisition, was published by Routledge in 2014.
Thursday, October 17th, 2019,
Jay Marciano:
Future Tense: How Artificial Intelligence and Neural Machine Translation are shaping our industry
Future Tense: How Artificial Intelligence and Neural Machine Translation are shaping our industry
the talk will take place, Oct. 17th, 2019, 3:00 p.m. –4:00 p.m., Business Administration Bldg, room 204 (BSA 204)
The confluence of improving translation technologies (better machine translation, smarter translation memory, cleverly automated workflows) and the quickly emerging world of Artificial Intelligence (including machine learning and Big Data analytics) are changing the language services industry and the day-to-day work of translators. Where is it all leading? What will the translation industry look like as software developers and LSPs continue to refine their tools? How will these technologies change the job of professional translator? This presentation will make informed predictions about what the future of these technologies will hold for professional translators, interpreters, and the translation industry over the next few years and beyond.
About JAY MARCIANO:
Jay Marciano, Director of Machine Translation at Lionbridge Technologies since 2010, has spent more than 20 years developing and applying MT. He serves on the Board of Directors of the Association for Machine Translation in the Americas. Prior to joining Lionbridge, he worked for SDL (2001-2010), where he was responsible for the development and refinement of natural language processing software. He has also worked as a product manager at Transparent Language, as a lecturer in the English Department at the University of Bonn, Germany, and as an editor at Houghton Mifflin on the staff of the American Heritage Dictionary
September, 6th, 2019
the talk will take place, 10:00 a.m. - 11:00 a.m. in the Business Administration Bldg, room 209 (BSA 209)
The application of neural network technology (“deep learning”) to Automated Speech Recognition (ASR) has had dramatic impacts on the quality of speech recognition systems. Seide et al 2011 showed, for instance, a 32% reduction in Word Error Rate (the amount of error when producing transcripts) over the previous state of the art, with no changes in training data. Up to that point, a drop of 1 or 2 percent was considered cause for celebration. More recent work in applying neural models to Machine Translation (MT) has shown equally dramatic improvements (e.g., Cho et al 2014, Devlin et al 2014, etc.). Most recently, Hassan et al 2018 achieved human parity quality on MT for news-specific test data using a combination of neural models. Improvements in both ASR and MT have increasingly made speech translation a viable technology, especially in highly technical discourse. For instance, we have found it of significant utility in technical talks, in the classroom and in lecture settings. In this talk, I will quickly review Microsoft Translator's work in this space, and will integrate a demo of our speech recognition and translation technology directly into the talk, so that audience members can follow along from their own devices.
About Will Lewis:
William Lewis is Principal PM Architect with the Microsoft Translator team at Microsoft. He has led the Microsoft Translator team's efforts to build Machine Translation engines for a variety of the world's languages, including threatened and endangered languages, and has been working with the Translator team on Speech Translation. He has been leading the efforts to support the features that allow students to use Microsoft Translator in the classroom, both for multilingual and deaf and hard of hearing audiences. Before joining Microsoft, Dr. Lewis was Assistant Professor and founding faculty for the Computational Linguistics Master's Program at the University of Washington, where he continues to hold an Affiliate Appointment, and continues to teach classes on Natural Language Processing. Before that, he was faculty at CSU Fresno, where he helped found the Computational Linguistic and Cognitive Science Programs at the university. He received a Bachelor's degree in Linguistics from the University of California Davis and a Master's and Doctorate in Linguistics, with an emphasis in Computational Linguistics, from the University of Arizona in Tucson. In addition to regularly publishing in the fields of Natural Language Processing and Machine Translation, Dr. Lewis is on the editorial board for the Journal of Machine Translation, was on the board for the Association for Machine Translation in the Americas (AMTA), served as a program chair for the National American Association for Computational Linguistics (NAACL) conference, served as a program chair for the Machine Translation Summit, regularly reviews papers for a number of Computational Linguistic conferences, and has served multiple times as a panelist for the National Science Foundation.
Spring 2019
May, 3rd,
The wide access to large data sets, compute power (GPUs), and deep learning has accelerated advances in multilingual natural language processing. The Natural Language Processing (NLP) field is growing substantially! I will discuss some of the most recent advances in language translation and the customization of language translation systems to new domains and languages. Given the richness of human languages, I will also present some the challenges that current NLP models fail to address adequately and are needed for Mastering Language by current NLP systems.
About Salim Roukos:
Salim Roukos is IBM Fellow and CTO for Translation Technologies at IBM T. J. Watson Research Center. Dr. Roukos received his B.E. from the American University of Beirut, in 1976, his M.Sc. and Ph.D. from the University of Florida, in 1978 and 1980, respectively. He joined Bolt Beranek and Newman from 1980 through 1989, where he was a Senior Scientist in charge of projects in speech compression, time scale modification, speaker identification, word spotting, and spoken language understanding. He was an Adjunct Professor at Boston University in 1988 before joining IBM in 1989. Dr. Roukos has served as Chair of the IEEE Digital Signal Processing Committee in 1988.
Salim Roukos currently leads a group at IBM T.J. Watson research Center that focuses on various problems using machine learning techniques for natural language processing. The group pioneered many of the statistical methods for NLP from statistical parsing, to natural language understanding, to statistical machine translation and machine translation evaluation metrics (BLEU metric). Roukos has over a 150 publications in the speech and language areas and over two dozen patents.
Roukos was the lead of the group which introduced the first commercial statistical language understanding system for conversational telephony systems (IBM ViaVoice Telephony) in 2000 and the first statistical machine translation product for Arabic-English translation in 2003. He has recently lead the effort to create IBM's offering of IBM Watson Language Translator and Watson Natural Language Understanding for building custom domain models for NLU applications in over a dozen languages.
March, 22,
For many migrants who perform vital services within the global economy, translation services are rare or non-existent. It is common for labor migrant to be forced to sign contracts in unfamiliar languages without fully understanding their contents. Many do not speak the languages of their host countries enough to navigate their complex legal systems, and there is little opportunity to learn them given the long hours worked and their requirement to live isolated from the local population. The limited availability and poor quality of translation resources suggests that, despite some progress, quality translation provision is not fully comprehended as integral to human and labor rights. Transnational identities associated with globalization are unavailable to them, as are the translation resources that would give them the information necessary to act in their own best interests. These migrants understand very well the significance of translation and its absence; they are actively involved in calling for improvements in the quality of translation services available to them. My talk considers a type of activism undertaken by some domestic and sex workers that takes the form of self-translation through art and poetry. In addition to highlighting their work conditions, these cultural productions allow them to give expression to their own lives and present themselves not as victims but as creative, multi-faceted human beings.
About Moria Inghilleri:
Moira Inghilleri is Professor of Translation and Interpreting Studies and the Director of Translation and Interpreting Studies. She is the author of Translation and Migration (Routledge 2017) and Interpreting Justice: Ethics, Politics and Language (Routledge 2012). She was co-editor of The Translator from 2011-2014 and review editor from 2006-2011 and served as series co-editor for the Routledge series New Perspectives in Translation and Interpreting Studies. She guest-edited and contributed articles in two special issues of The Translator: Bourdieu and the Sociology of Translating and Interpreting (2005) and Translation and Violent Conflict (2010, co-edited with Sue-Ann Harding). In 2017 she was appointed to the Fulbright Specialist Roster for 2017-2020 in the field of translation and migration studies.
February, 22,
... On the one hand, machine translation (MT) systems are "real" AI: they ingest documents written in one language and spit out translations 24 hours a day, on any topic you want, with never a human translator in sight. And they've done so since the 1970s, with ever-increasing (and still unpredictable) accuracy. On the other hand, there are "translation memory" (TM) systems that augment translators' intelligence, or at least their memory. These TM systems store thousands of sentences with their human-created translations and for any new sentence they find the best suggestions among those stored translations. Human translators choose from and adapt these suggestions to create their final translations. Clearly, translation memory systems don't do much at all without human users. ...
What could be in the middle between these total opposites: machine intelligence and human expertise? What's in Markoff's "common ground between humans and robots"?
The middle is where you'll find the 50 shades of AI: the many ways to mix and match machine intelligence with varying amounts of human expertise. These are the options and opportunities that the AI dichotomy makes us blind to; the very many possibilities besides the focus on human-free intelligence that seems to have hypnotized and horrified so many people ....
(complete text: https://www.linkedin.com/pulse/50-shades-ai-mike-dillinger-phd/)
About Mike Dillinger:
Dillinger, PhD, is Manager of Taxonomies and Human Judgements in the AI Division at LinkedIn where his team teaches machine learning algorithms about the world of work. Before that, he was Technical Lead for eBay’s and LinkedIn's first machine translation systems, an independent consultant specialized in deploying translation technologies for Fortune 500 companies, and Director of Linguistics at two machine translation software companies where he led development of the first commercial MT-TM integration. He was President of the Association for Machine Translation in the Americas and has two MT-related patents. Dr. Dillinger has also taught at more than a dozen universities in several countries and has been a visiting researcher on four continents.