TAUS - Enabling better translation

TAUS USER CONFERENCE 2011
Santa Clara (CA), USA / October 6 - 7

---

---

  

---

PROGRAM

Wednesday, October 5

---

16:30 – 18:00 / TDA General Assembly Meeting Agenda:

• Amendments of the Member Regulations
• Approval of the Annual Accounts
• Budget Review
• Board Members Election

CLICK ON EACH PRESENTATION TO VIEW IT ON YOUTUBE.

Thursday, October 6

---

08:30 / REGISTRATION

09:00 / Welcome and program overview
Jaap van der Meer, TAUS

 

THE MULTILINGUAL WEB

9:15 / Keynote Panel
A panel of thought leaders debate the next phase of the multilingual web - beyond the ‘translate’ button.

Panelists:
Bruno Fernandez Ruiz, Yahoo! Fellow and Vice President
Bill Dolan, Head of NLP Research, Microsoft
Addison Phillips, Chair, W3C Internationalization Group / Globalization Architect, Lab126

Moderators: Rose Lockwood and Jaap van der Meer, TAUS

10:15 / REFRESHMENT BREAK

10:45 / TAUS Program Management Office
Jaap van der Meer, TAUS
An overview of the objectives, ways of working and progress of the TAUS PgMO

 

INTEROPERABILITY

“We envision translation as a standard feature, a ubiquitous service. Like the internet, electricity, and water, translation is one of the basic needs of human civilization.”

TAUS Mission statement

Interoperability is a pre-requisite to deliver on this vision and mission. The industry’s lack of interoperability is costing a fortune and hindering innovation. However, there is a revived surge towards openness and sharing - key criteria for change. The Interoperability block includes:

1. Open Translation Platforms
2. Open Technology Exchange
3. The Great Interoperability Debate

11:00 – 11:45 / Open Translation Platforms
In 2009 TAUS introduced the theme 'Open Translation Platforms' to its events. Each subsequent event has given special focus to specific aspects of open platforms. The following three presentations were chosen for their potential for cross industry impact

11:00 / Building a common RESTful API for translation services
Brian McConnell, Worldwide Lexicon

Based on an upcoming TAUS report, this presentation describes how the RESTful API can be applied to translation services to create a common interface, yet without requiring individual developers or vendors to make fundamental changes to the way their component operates. This will be easy and inexpensive to implement and will be easy to build applications around.

11:10 / XLIFF 2.0: Great (Processing) Expectations
Andrew Pimlott, Welocalize

We explore some of the difficulties we encountered when enhancing XLIFF support in the open-source OpenTM2 translation workbench and present some proposed processing expectations based on the development version of XLIFF 2.0. These will demonstrate ease of implementation while increasing interoperability.

11:20 / OAXAL - Open Architecture for XML Authoring and Localization
Robert Willans, XTM International

Robert Willans outlines a reference implementation of OAXAL, which is an OASIS TC Reference Architecture standard incorporating all of the relevant localization standards in one SOA (Service Oriented Architecture). OAXAL provides a template of how to achieve a high level of automation in the authoring, localization and publishing cycle using Open Standards throughout.

11:30 / Q&A with speakers and invited experts
Brian McConnell, Worldwide Lexicon; Andrew Pimlott, Welocalize; Robert Willans, XTM International; Daniel Gervais, MultiCorpora; Chase Tingley, Welocalize

CLICK ON EACH PRESENTATION TO VIEW IT ON YOUTUBE.

11:45 – 12:30 / Open technology exchange

TAUS is launching TAUS Labs, a platform for the exchange of open-source technology and the development of knowledge tools for TAUS members. A number of TAUS members have developed tools and scripts that they are sharing to help others. In the same spirit, TDA has opened up the super cloud of translation memories for everyone to use. The session introduces TAUS Labs and a selection of open tools presented by their developers.

11:45 / Introducing TAUS Labs
Rahzeb Choudhury and Maxim Khalilov, TAUS

11:55 / AIR-based light clients for supporting Moses engine training
Jeff Rueppel, Adobe

Within Adobe we have recently begun training statistical machine translation engines using the open-source Moses package. In order to make engine creation simpler, we have developed a set of supporting tools. Each AIR-based client is lightweight and features an easy-to-use and configurable interface to automate specific processing tasks. The tools created include:

• a corpus-cleaning tool that translates TMX files into a Moses-ready format
• a training harness that simplifies the kick-off of engine training
• a scoring harness that calls a number of automated scoring routines

12:05 / Simple Table Editor for TMX files for Moses
Leonid Glazychev, Logrus

Data-driven linguistic projects using TDA’s repository require large amounts of past translations processed to train MT systems such as Moses, as well as other tools and systems. The editor is made available as free software (without the sources) to TAUS, TDA, Moses and other professionals in the localization industry.

12:15 / Moses for Localization – Real world integration
Achim Ruopp, Digital Silk Road

You trained a customized Moses MT engine – now what? The engine needs to be integrated into localization workflows for post-editing and content management systems for gisting machine translation (MT). We present two real-world integration case studies:

• Segment MT in an existing TMS via a client-server API
• On-demand document MT via a SOAP web API

Learn how the open-source project Moses for Localization provides a toolbox for typical integration scenarios.

12:23 / SymEval – A GTM based evaluation
Fred Hollowood, Symantec

An update on the SymEval MT output quality evaluation tool which is now available on SourceForge.net.

12:30 – 14:00 / LUNCH

14:00 – 17:30 / The Great Interoperability Debate
Session led by Greg Oxton, Consortium for Service Innovation

TAUS has launched the PgMO to represent members and provide strategic change leadership, improved governance and enhanced execution of the industry’s interoperability initiatives. However, key questions remain unanswered from previous “Great Interoperability Debates.” How do we solve the interoperability problem? What are the business drivers that will foster greater interoperability? How do we break away from the complexity that has been built (or that we built) around us? How do we work together to meet the interoperability challenges that lie ahead? What do we tackle now? What comes next? For this final “Great Interoperability Debate” in 2011, we give the entire afternoon and floor to all TAUS User Conference participants and ask them to help answer these and other burning questions. We use the tried-and-tested Open Space Technology format. The session will be led by Greg Oxton of the Consortium for Service Innovation. The interoperability debate can zoom in on CMS and localization/language technology interoperability; interoperability challenges and opportunities with social media and social media management; multiscreen and multimedia publishing – in fact, whatever participants deem as relevant.

16:30 / Report back to plenary

17:30 / CLOSE

18:30 / NETWORKING RECEPTION AND DINNER
Magnolia Room, Lobby Level, Hyatt Regency Santa Clara

CLICK ON EACH PRESENTATION TO VIEW IT ON YOUTUBE.

Friday, October 7

---

07:30 / CALL FOR IDEAS – BREAKFAST BRAINSTORMING
All conference participants are invited for this early morning session to put forward ideas on how to improve the TDA service offering.

09:00 / Welcome back and day overview
Jaap van der Meer, TAUS

TRANSLATION AUTOMATION

The morning session is devoted to announcements of MT breakthroughs and use cases, as well as talks about multilingual interaction applications for social media and customer support. We ask leading researchers to look forward and share their insights in the possible convergence of TM and MT and ways to better handle domain adaptation in MT training.
On the horizon are a new generation of tools that integrate statistical methods and advanced linguistic intelligence, and apply these techniques to very large corpora. How will today’s research find its way to new products and use scenarios? What will be the impact on productivity and on translation strategies?

New Breakthrough in MT

09:15 / Introducing ABBYY Compreno – new approach to machine translation
Eugene Pakhomov, ABBYY

ABBYY Compreno is the new, intelligent MT system, which provides unequalled quality of translation from Russian into English and from English into Russian. The system is a product of many years of research in natural language processing and analysis and is capable of analyzing the syntactic and semantic structure of texts.

MT USE CASES

9:45 / Letting 1,000 MT systems bloom
Daniel Marcu, SDL Language Weaver

Last year, TAUS challenged the industry to “Let 1,000 MT systems bloom.” SDL took this challenge seriously, and in this presentation, Daniel Marcu, CTO of SDL Language Gateway and founder of SDL Language Weaver, will focus on the lessons learned along the way. He will also shed light on the science and business implications that follow from the experience gained in building more than 10,000 MT engines.

10:05 / REFRESHMENT BREAK

10:35 / Quality above price and speed
Andy Way, Applied Language Solutions; Catherine Dove, PayPal; Donald Johnson, Caterpillar

Two use cases to highlight how quality has emerged as a key reason for organizations to move to MT, based on results of post-editing customized MT output. Andy Way will help explain the reasons behind the results now being achieved, which are surpassing human translation in many cases.

11:05 / Rapid Development and deployment of client-adapted Nordic MT engines for PayPal
Olga Beregovaya, Welocalize; Catherine Dove, PayPal; Julia Epiphantseva, PROMT; Alon Lavie, Safaba Translation Solutions

A joint presentation highlighting the successful results of technological collaboration between PROMT and Safaba to deliver on PayPal’s requirements, with linguists at Welocalize and Paypal providing testing and feedback.

CLICK ON EACH PRESENTATION TO VIEW IT ON YOUTUBE.

MULTILINGUAL INTERACTION USING MT

The digital space is increasingly real-time and multimedia (from text to speech). In this session we learn from three use cases of multilingual interaction applications for social media (communities, discussion boards, blogs), multilingual chat, IM, smart phones.

11:20 /Using MT for real-time multilingual interaction and collaboration
Will Burgett, Intel

The potential for using MT to enable multilingual interaction and collaboration may dwarf its use in the more traditional areas of document, web and software translation. Integration with applications such as chat, IM, email, blogs, forums, speech-to-speech and collaboration software makes MT as limitless as the number of opportunities for human-to-human interaction. This presentation will review some key results, discoveries and challenges of using MT integrated with chat and community forums to enable multilingual communications.

11:35 / Microsoft MT for IM bot and Lync
Chris Wendt, Microsoft

In this talk, Chris Wendt outlines the use of IM bot for Messenger and a translation plugin for Lync. He describes a set of analysis tools that are used to provide a clear picture of market perception for Microsoft products. He looks at using automatic translation for the inbound translation; and individually crafted messaging for the outbound portion.

11:50 / Asia Online and multilingual chat
Kirti Vashee, Asia Online

Asia Online has developed a cross-language chat capability that enables customer support professionals to quickly understand multilingual questions and respond with carefully prepared responses in multiple languages. This talk will overview some early trials and explore how this can further evolve to monitor and engage with customers in “internet time” wherever brand impacting conversations are being conducted.

12:05 / Real-time multimedia translation for Healthcare: A pilot project
Mark Seligman, Ph.D., Spoken Translation, Inc.

This presentation reports on a pilot project applying Converser for Healthcare, Version 3.0, a real-time, multimodal, broad-coverage, highly interactive translation system, in three departments of a large hospital complex belonging to a major US healthcare organization – pharmacy, inpatient nursing, and eye care. We survey lessons learned concerning software; use cases and setups; equipment and technology; and logistical and processing issues.

12:20 / Q&A with session speakers

12:40 – 14:00 / LUNCH

CLICK ON EACH PRESENTATION TO VIEW IT ON YOUTUBE.

RESEARCHERS DEBATE ON FUTURE TRANSLATION TECHNOLOGIES

We are often using pretty much the same technologies. Nobody is getting results that are head and shoulders above the competition. Researchers focus on a myriad of nuances in search of improvements. There are high hopes with hybridization, customization and syntax, among others. And then there are many open questions about the role and efficient use of data. Talks from two major research groups help to ground us in reality, and a panel of the leading researchers join the debate to help shed light on what we can expect in the near future.

14:00 / MT is here. Where is it going next?
Chris Wendt, Microsoft
A down-to-earth and forward-looking talk about the five things that move the landscape: (1) the enormous growth in the extent and type of translatable material; (2) the interaction with the MT system itself; (3) the use of TMs; (4) the effect of training data; and (5) system integration.

14:15 / MT at the CNGL
Declan Groves, CNGL

The CNGL MaTrEx machine translation framework represents over a decade’s worth of state-of-the-art research in MT carried out at DCU. In this talk we outline some of the innovative MaTrEx components that have been identified as having potentially high commercial benefit, including dynamic domain-adapted MT, components for the integration of existing in-house TMs with MT and novel methods for improving the quality of the training data we use.

14:30 / Q&A with session speakers and invited experts
Chris Wendt, Microsoft; Declan Groves, CNGL; Daniel Marcu, SDL Language Weaver; Andy Way, Applied Language Solutions; Alon Lavie, Safaba Translation Solutions 

15:00 – 15:30/ REFRESHMENT BREAK

INNOVATION

Interoperability, open translation platforms and automation lead to great opportunities for innovation. In this final session, we will look at new business models and the convergence of technologies leading to exciting new offerings in the translation industry. Opportunities for innovators also create threats for established players.

15:30 / Microsoft’s Collaborative Translation Framework
Vikram Dendi, Microsoft

15:50 / Catalyzing business model innovation
István Lengyel, Kilgray Translation Technologies

MT post-editing is often paid by the hour. The problem with hourly rates is that translators tend to have different speeds. Kilgray’s memoQ 5 allows us to quantify changes, and this proposes a new business model: pay some checking fee per word or character, and pay some editing fee per words edited. By knowing what the post-editor has changed, translation industry players can also know what should be improved in MT, what should be entered into the dictionary, and what sort of sentences need to be trained.

16:10 / CLOSING

16.30 / ADJOURN