QTLaunchPad Showcase

GALA is pleased to partner with QTLP to highlight their work on industry quality metrics

QTLP

Building a User-Centric Translation Quality Assessment Model: The QTLaunchPad Project

High-quality translation is in greater demand today than ever before: the lack of fast, affordable quality translation hurts business and communications. Despite the productivity gains enabled by translation memories, workflow improvements, and other technologies, human translation has reached its limits. As Paula Shannon from Lionbridge puts it, we have reached the stage where translation is “beyond human scale.” As a result, the volume of translation today falls far short of what is needed for optimal business operations and legal requirements. Industry, administration, and society all urgently need progress in novel translation technologies and workflows that integrate human and machine translation.

Unfortunately translators must spend significant amounts of time (up to 50% according to one LSP’s internal studies) on routine and fundamentally non-productive tasks. When translators are freed to focus on creative, challenging, and interesting tasks while machines handle routine and repetitive tasks, we will be able to extend multilingual communication to additional languages and services (e.g., to conquer untapped markets) and reduce costs associated with commitments to linguistic diversity, all while improving the status of human translators. The QTLaunchPad project, which has organized a workshop at GALA 2013, aims to improve the standing of human translators and the quality of the technologies and resources that form an essential part of their working environment.

Today’s Machine Translation Cannot Meet the Need for High Quality

Although the quality of market leaders for automatic gist translation keeps improving, their output is still unusable for most outbound translation requirements, even as a source for cost‐effective post‐editing. Today, nearly all on‐going machine translation research is dedicated to gradual improvements of statistical gist translation. While large, gist translation services do not offer business opportunities for small- and medium‐sized enterprises, quality translation technology actually does: it requires strong specialization and close cooperation between technology providers, translation service providers, and corporate customers.

The Workshop

Translation Quality Assessment (TQA) is an important business topic. Formal metrics such as the LISA QA Model and SAE J2450 for human translation have helped, but metrics for machine translation (e.g., NIST, BLEU, or METEOR) are currently suitable in very limited scenarios: to compare variants of a given translation system or measure system progress over time, not for production environments. Buyers have no way to make meaningful comparisons of the quality from different solutions and providers or to tell if quality levels meet their actual needs. Current TQA has a one-size-fits-all approach: gist translations, regulatory and legal documents, and marketing material are all evaluated as though they were the same. The QTLaunchPad project has developed an open and unified “multidimensional” framework to TQA (launched publicly in early 2013) that is built around quality metrics. The metrics are complementary to efforts such as the TAUS Dynamic Quality model (and provide “hooks” to include data from other models) and can precisely replicate the functionality of existing models such as the LISA QA Model or SAE J2450 for users who wish to maintain compatibility with legacy efforts while still taking advantage of new features. In addition to richness and compatibility, one of the central design features of the framework is usability.

This workshop (Tuesday at 16:00) will focus on the measurement of translation quality. It will introduce attendees to the metrics mentioned above. It will demonstrate tools for creating project type-specific metrics and ensuring their validity for actual production tasks. Our goal is to get participants to provide feedback and to discuss their own quality requirements and needs to help improve and further develop the Multidimensional Quality Metrics model. The workshop will feature hands-on, discussion-oriented exploration of key issues related to TQA.

Get Involved

As the QTLaunchPad project progresses, results and information are made available for public discussion and feedback via www.qt.21.eu/launchpad. To be added to our announcement mailing list or for inquiries, please contact [email protected].

About QTLaunchPad

QTLaunchPad is an EC-funded collaborative research initiative dedicated to overcoming quality barriers in machine and human translation and in language technologies. In addition to the quality metrics mentioned above, it is preparing a large-scale translation quality initiative for Europe. The consortium consists of four world-leading research centers:

QTLaunchPad is collaborating closely with GALA and other organizations and stakeholders in the language services industry to ensure that its resources are widely usable and appropriate for meeting a variety of needs.

Special Events

“The GALA atmosphere encourages an incredibly high level of frankness and networking.”

Stanislav Kalianov, Schneider Electric

Last Year at GALA 2012

 
77%  reported making contacts that may generate new business for them.

84%  said they learned new concepts or strategies that will help them in their business.

95%  reported that they were satisfied or extremely satisfied with the conference.

 

What's New

Why Attend?

See what past attendees value about GALA Language of Business conferences.


Conference Testimonials