TRANSLATION IN THE 21ST CENTURY
TOKYO, APRIL 19 – 20, 2012 (HOSTED BY ORACLE JAPAN)
TAUS Executive Forums are two-days meetings for buyers and providers of language services and technologies, aimed at an open exchange about language business innovation and translation technology. The TAUS Tokyo Executive Forum 2012 is the second TAUS meeting in Japan. The first took place in April 2010 and was also hosted by Oracle Japan.
TRANSLATION IN THE 21ST CENTURY
Language and translation business is undergoing dramatic changes, characterized by a growing number of languages, a shift to more dynamic content and user-centric publishing. In this rapidly changing environment buyers and providers of language services and technologies are challenged to change their business models. In the TAUS Executive Forums we invite language industry operators to help define strategies for translation in the 21st century.
Buyers of translation are in need of an enterprise language strategy to be ready for expansion in languages, growth in social media, rapid turn-around translation and effective global customer support in the cloud-computing age. Soon translation will be embedded in every application and service we use. This is very different from the current project-oriented approach to translation. How do we make this transition?
Providers of translations services and technologies need to redefine their business and value for their customers. Increasing efficiencies by introducing more technology is just part of the full picture. New opportunities are emerging for the most apt innovators.
TRANSLATION IN THE 21ST CENTURY
Topics at the TAUS Executive Forum are all part of our concept of “Translation in the 21st Century”. The key words are:
- Machine translation and translation optimization
- Collaborative translation and crowdsourcing
- Interoperability of translation tools and resources
- Automated real-time translation
- Spoken translation
- Language and translation business innovation
- Community/voluntary translation
- Translation memory and language data sharing
For background on these copies, please click here to read a series of published articles.
WHO IS ORGANIZING
The TAUS Executive Forum is organized by TAUS, the Translation Automation User Society. TAUS is a think tank for the translation industry, undertaking research for buyers and providers of translation services and technologies. Our mission is to increase the size and significance of the translation industry to help the world communicate better. To meet this ongoing goal, TAUS supports entrepreneurs and principals in the translation industry to share and define new strategies through a comprehensive program of events, publications and communications.
The program for the TAUS Tokyo Executive Forum is created and reviewed by a Program Committee of experts and leaders in the field:
- Professor Hitoshi Isahara, AAMT
- Hiroki Kawano, Honyaku Center
- Junichi Chigira, Oracle
- Hirokazu Suzuki, Toshiba
- Tetsuzo Nakamura, Yamagata
- Hideo Yanagi, Acrolinx
- Professor Sadao Kurohashi , Kyoto University
- Professor Eiichiro Sumita, NICT
- Professor Kyo Kageura, Tokyo University
Thursday April 19
Part 1: Vision & Market Overview
Junichi Chigira (Oracle)
9:10 /Agenda overview & introductions
Jaap van der Meer (TAUS)
9:20 /Who Gets Paid for Translation in 2020
Jaap van der Meer (TAUS)
Google, Microsoft, Baidu, Yandex and Yahoo! are either getting paid or getting ready to be paid for translation through advertising revenue. Giving access to multilingual information increases their user base and raises cash. They are setting an example that many others would like to emulate. Anyone who fails to see the fundamental shift in the demand for translation from the traditional buyer to the billions of citizens, patients, tax payers and consumers, is just scratching the surface of the vast potential for the global language industries.
The European Commission tells us that each EU citizen is paying on average €2 per year to fund the one Billion Euro translation budget of the Directorate General of Translation, by far the largest in the world. We can rightfully say that translation is already being paid for in different ways than the word-price model.
Data-driven machine translation is still in its infancy and the language industry is only just starting to work with this technology on a material scale. We are only at the beginning of an innovation journey that will include game changing shifts for both buyers and providers of translation as they seek to adapt their models to the 21st century. In his opening presentation Jaap van der Meer will give an overview of “Eight Things to Change”.
9:45 /New MT developments in Asia
Hitoshi Isahara, President (Asia-Pacific Association for Machine Translation)
In this talk, Professor Hitoshi Isahara will share an overview of current activities in MT in both research and commerce in Japan and in Asia. He will also introduce several basic tools to support human process of translation, such as term extraction and equivalent selection.
10:10 /Machine Translation and Digital Libraries
Makato Nagao, President (National Diet Library)
Machine translation has been developed for more than forty years, and is now used in varieties of real tasks although it is not fully satisfying. This talk provides an overview of a multilingual user interface for library information of Japanese, Korean and Chinese National Libraries through machien translation.
10:30 /Refreshment break
11:00 / Changing Our MT Perspective
Daniel Marcu (SDL)
In 2002, statistical machine translation was commercialized to bring a new translation solution to enterprise and government organizations. While this was a great advance for global communication, it has also stirred a decade long debate over which systems are the best. First it was Rule Based MT vs. SMT, then RbMT vs. Hybrid vs. SMT. Today, the conversation continues about which individual advances or improvements generate the best translations. It is time to change our perspective.
In this presentation, Daniel Marcu, Chief Technology Officer for SDL Language Technologies will share his vision for automated translation. He will talk about what is required to transition MT from a component into an ecosystem - one where MT is touched by numerous products and platforms and becomes an integral part of each and every communication. In this type of environment, improvements to systems drive significant productivity gains across the board, give users control over the MT systems for their particular communication needs and create opportunities for both the enterprise and translation vendors. The outlook is bright, but it starts with changing the conversations we are driving as a community.
Part 2: Best Practices in Translation Automation
11:30 /Use case Fuji Xerox
11:55 /MT plus post-editing in an English-to-Japanese localization context: How useful can it be, compared to translation memory, and how does it change proessional translators' production style?
Masaru Yamada (Rikkyo University)
The aim of this presentation is to uncover how the integration of a machine translation plus post-editing (MT post-editing) in an English-Japanese translation setting will change professional translators' translating process and performance. When MT post-editing is implemented as a translation assistance strategy, human translators are forced to concentrate on revising the MT output.
This new form of translating activity in turn brings up a question that concerns the acceptance of machine-translated texts, and how much human effort is necessary to improve such imperfect texts. This question of acceptance of MT output has been evaluated through empirical researches in terms of productivity gain and compared to the threshold point at which the raw MT output will (not) support increased translation speeds, as often seen with Translation Memory low fuzzy-match operations. These results are presented and discussed in the presentation.
12:20 /Discussion & Questions
14:00 /MT:Pushing the limits
Catherine Dove (PayPal)
We have all heard of MT as a tool to increase translation turn-around time and lower translation costs. While MT stands for Machine Translation, is is More than That. In this study, Paypal will share their experience of using MT to ensure that all new designs and contents are L10N-ready prior to translation start, achieving the biggest gain of all int heir global SimShip environment.
14:30 /The Utilization of Machine Translation Systems in the Japan Patent Office
Mr. Eiichi Yamamato, Deputy Director, Patent Information Policy Planning Office (Japan Patent Office, JPO)
Some large patent offices utilize machine translation to get the gist of the documents written in foreign languages and offer its patent information to other countries. Mr. Yamamoto will introduce the JPO efforts on machine translation such as cooperation with other offices, efforts to improve machine translation quality, and future plans.
15:00 /Simplified Japanese for Enhancing Translatability
Midori Tatsumi (Technical University of Toyohashi)
Report on the feasibility of 'simplified' Japanese adequate for both user manuals and documents capturing the know-how of key employees to share with operations overseas. This presentation is based on a joint research project by 'Professors Tony Hartley, Midori Tatsumi, Hitoshi Isahara (Toyohashi University of Technology), Kyo Kageura (University of Tokyo)'. In their experiments they tested various writing rules, eliciting judgments of their felicity in Japanese and the quality of their English translations by several MT systems.
15:20 /Discussion & Questions
15:30 /Refreshment break
16:00 /Leveraging Moses in Oracle-proprietary translation factory
Hiroko Matano and Reiko Saito (Oracle)
In Oracle, we have integrated Moses MT engine into Translation Factory, our mission critical translation workflow system, and started to use MT with post-editing for certain production projects in 4 European languages and Japanese. This presentation shares our approaches and experiences for Japanese on 1) how our MT translation workflow system works, 2) how tagged files are processed in Moses, and 3) how we support post-editing and control its quality.
16:30 /Findings from English to Japanese Post-Editing Productivity Tests Using Three Different Moses Configurations
Mirko Plitt (Autodesk)
Autodesk uses Moses to translate software strings and product documentation from English to Japanese, among other languages. All MT output undergoes full post-editing. Before putting the Japanese Moses engine to production, Autodesk carried out two post-editing productivity tests involving Japanese. The first one, using a standard Moses process, did not show any productivity gain, whereas the second, including among others an engine integrated in a custom-developed reordering process for the English source text, showed sufficient productivity gains to justify going to production.
The presentation will provide a few insights into the findings made during the Japanese productivity tests. This will also include preliminary results from a study underway at Geneva university to identify the correlation between Japanese post-editing productivity and "Post-Editing Actions", a more sophisticated approach to measuring the changes carried out by post-editors than the traditional edit distance metrics
17:00 /Discussion & Questions
18:30 /Networking dinner at Restaurant Akasaka Square Dining on the 3rd floor of the Akasaka Excel Hotel Tokyo (few minutes walk from the New Otani hotel) (Confirmation required)
Friday April 20
Part 2 continued: Best Practices in Translation Automation
9:00 /Introduction of the day
Jaap van der Meer (TAUS)
9:10 /Lessons learned from Honyaku Center's experience of Having Original CAT Tools
Hiroki Kawano (Honyaku Center)
Honyaku Center is the largest and only public translation provider in the Japanese market. Since going public in 2006, the company has planned to prepare an original CAT environment to increase its competitiveness. Hiroki Kawano will introduce what the company has experienced and learned from their effort in these years, and where they want to go from here.
9:35 /MT Post-Editing for High Volume TechPub
Alan Chung (SDL) and Dieu Tran (Cisco)
Joint Presentation from Cisco Japan and SDL Japan. We will present how Cisco operationalized a MT post-editing solution to reduce cost, and cycle time for localization of large volume product manuals. We will describe approach used to train a hybrid engine, up to actual production rollout and continuous customization, involving a close collaborative effort between MT vendor, SDL Japan and Cisco Japan.
10:00 /Discussion & Questions
10:10 / Refreshment break
Part 3: Innovation Use Cases
10:30 /Crowd and Computer Translation Collaboration
Miori Sagara (Baobab)
[Ryugakusei Network @ Minnna no Hon'yaku] is a new online crowd-translation platform that allows people in the crowd to work with and build from statistical machine translation, with unique, gamification tools.
10:55 /Multilingual Chat for Support at Honda
Geert Benoit and Heidi van Hiel (Yamagata)
Honda Motor Europe is working with Yamagata Europe on two MT projects to support the Service and Repair division in its transformation from a decentralised model to a centralised service model. Both projects relate to the repair of a broad range of Honda products (Car, Bike and Power Equipment) and share terminology and MT engines.
11:20 /Collaborative Machine Translation: Your community and Microsoft's knowledge working together
Chris Wendt (Microsoft)
The implementation of collaborative translation, in conjunction with personalized training of the translation engine, provides companies and interested groups a unique and elegant option of using their own data, and the feedback of internal and external crowds, to tune and improve a translation engine for specific uses - or for new languages. Big data from Microsoft's Bing Search engine provides generic language knowledge, while the translations and target language material provided by your community provides the specific domain knowledge: trained and optimized together gives you the best of both worlds, and significantly improves the utility of machine translation in your field, above a generic engine. We will look at the technology involved, and a few practical examples of the collaboration and custom training working together, lifting the quality of automatic translation.
11:45 / Multilingual Communication Platform - WorldJumper
Suguru Sakanishi (Yaraku, Inc.)
Worldjumper.com is a multilingual communication platform which translates users' post into other languages with auto-translation and social translation. It provides several translation solutions:
- WorldJumper Contact Form: with the solution, clients can communicate with foreign customers in their native language. (Japanese) when they send a message through a contact form on a corporate web site, the message will be translated into Japanese and the officers can reply and vice versa.
- WorldJumper Facebook App: The app makes the Facebook wall multilingual. When a page administrator posts through the app, it will be in 10 languages on the wall and appear in the fan's newsfeed in their language (if the language setting of the fan is Spanish, the post is showed in Spanish in his feed).
12:10 /Translations by the hour - a race to the top for the translation industry
Grant Straker (Straker Software)
This presentation will look at the "race to the top" for the translation industry. Who can translate the fastest, with the best quality and at the best price. Straker has developed a world leading "translations by the hour" system which changes the way in which translation projects are managed and charged. Find out why this new model is a game changer for the translation industry.
Part 4: Translation Quality Evaluation
14:00 /TAUS Dynamic Quality Framework
Rahzeb Choudhury (TAUS)
Quality is when the buyer or customer is satisfied. Yet quality measurement in the translation industry is not always linked to customer satisfaction, but rather is managed by quality gatekeepers on the supply and demand side who have specific evaluation models, the majority of which are based on counting errors, applying penalties and maintaining thresholds with little, if any, interaction from customers. Quality evaluation (QE) in the translation industry is problematic. Despite very detailed and strict error-based evaluation models, it seems that satisfaction levels with both translation quality and the evaluation process itself are low.
QE models are static, that is, there is a ‘one size fits all’ approach. Little consideration is given to multiple variables such as content type, communicative function, end user requirements, context, perishability, or mode of translation generation (whether the translation is created by a qualified human translator, unqualified volunteer, machine translation system or a combination of these). Together with a group of twenty of the enterprise members TAUS is introducing the Dynamic Quality Framework, a knowledge base for general industry use, which will be followed by a Dynamic Quality Dashboard for QE benchmarking.
14:20 /MT Evaluation within Symantec
Kiyoshi Izumi (Symantec)
Symantec has been a long-time user of Machine Translation. In this presentation we learn about new approaches to evaluation MT output.
14:40 /Guidelines for Measuring and Comparting Machine Translation Output
Dion Wiggins (Asia Online)
MT can be measured for a number of purposes ranging from comparing improvements in quality between two versions of a custom engine to comparing the quality of different MT providers. It can also be measured against human only approaches to translation. This presentation will explore automated metrics and provide guidelines and best practice for comparison of MT that ensures that a meaningful set of metrics can be provided.
There are many automated metrics for machine translation, each with the benefits and drawbacks. When combined with human post editing metrics, an entirely new range of metrics are available for comparison of MT with human only approaches. We explore each metric with real world case studies.
15:00 /Authoring-translation integration use case for...
Hideo Yanagi (Acrolinx)
15:20 /Refreshment break
Part 5: Data is core
16:00 /The PangeaMT-Toshiba Connection: Tackling domain-specific and customized MT for Japanese-English using TAUS Data
Tokiharu Iwanaga (Pangeanic/BI Japan) and Hirokazu Suzuki (Toshiba)
This is a proof-of-concept experiment aimed at testing the viability of combining both companies’ MT approaches into one that, using vast amounts of in-domain data coming from TAUS, leads to a satisfactory hybrid system for the ENJP pair (both directions in mind) with a view to expanding to other language pairs entailing similar difficulties later on.
The first step has been for Pangeanic to provide Toshiba with enough corpus [bilingual data coming from TAUS Data] in the EN <-> JP combination, although we have prioritized work in the EN>JP direction first. Toshiba was then involved in the nipponization of the EN component, i.e. to process it in such a way that the EN is reordered in a JP word-order (WO) fashion.
This newly Nippon-EN then gets realigned with the genuine JP target, and the resulting bilingual data sets in TMX format become the training corpora for SMT at Pangeanic’s end. This language reordering according to the other language’s preferred word-order (WO) system has proven to be a useful pre-processing module in our framework to make the translation request’s input language closer in style and word-order to the target language.
16:20 /The Language Grid: Service-Oriented Multi-Language Infrastructure
Donghui Lin (Kyoto University)
The Language Grid is a service-oriented collective intelligence platform for language services. In the Language Grid, the language services are wrapped by language resources collected from companies, university laboratories and research institutes, including morphological analyzers, parallel texts, bilingual dictionaries, concept dictionaries, machine translators and so on. Currently the Language Grid is federatedly operated by Kyoto University, Japan and NECTEC, Thailand, with participation from more than 140 organizations among 18 countries and regions. We are also planning to expand the Language Grid in Asian areas. In this talk, we will introduce the architecture, operation and applications of the Language Grid.
16:40 /TAUS Data 2020
Jaap van der Meer (TAUS)
TAUS Data Association is an industry-owned platform for the sharing of language data. Members and non-members can upload and download translation data for the training of MT engines and for advanced leveraging. Translators use TAUS Data to search terms and phrases.
17:00 /Discussion & Questions
Next TAUS Tokyo Executive Forum: April 11 - 13, 2012 (hosted by Oracle Japan).
The TAUS Executive Forum will be hosted by Oracle Japan at the Oracle Aoyama Center, 2-5-8 Kita Aoyama, Minato-ku, Tokyo 107-0061 Japan. Main Phone: +81.3.6834.6666 . The Oracle Aoyama Center is directly connected from Gaien-Mae Station/Metro Ginza Line, Exit 4B. The reception desk is on the 2nd floor. It takes about less 15 min by taxi from hotels below. Click here to view the map.
Recommendable Hotel (5 star class):
HOTEL NEW OTANI TOKYO
4-1 KIOI-CHO, CHIYODA-KU, TOKYO, 102-8578 JAPAN
Tel : 81-3-3265-1111 Fax : 81-3-3221-2619
Registration fees for the TAUS Tokyo Executive Forum are:
- €500 for members of TAUS (appr. 50,000 JPY)
- €1,000 for non-members of TAUS (appr. 100,000 JPY)
TAUS offers a 50% discount on the annual membership fees for new organizations registering for the TAUS Tokyo Executive Forum and joining TAUS. This offer is limited to companies headquartered in Asia.