TAUS - Translation Automation

Crowd/cloud/community sourcing in the translation industry is getting much eyeball attention, but there’s a lack of detailed insight into what works and what doesn’t. The TAUS User Conference helped share useful practices by fielding use cases and a powerful overview of community dynamics that set the agenda for this very topical phenomenon.

Cisco has fully engaged with the potential for crowdsourcing some of its localization requirements by integrating the crowdsourcing process, together with a translation toolkit and MT on a central portal and handling it internally. The idea is to enhance the value chain, from authoring the source to translating and quality assessment, by offering an online platform.

Like many IT companies, it already has a crowdsourcing experience in the form of Support Wikis and professionals-only forums. But localization required a dedicated initiative: the content is complex, terminology needs to be respected, and the service has to mesh with the company’s operating environment.

To address vital quality and control concerns, the crowd had first to be identified and trusted, Cisco currently crowdsources from its regional support engineers around the world. They know about the products and they are comfortable with the English source.

The translation tool provides Cisco TM leverage and MT for segment translation, plus a more advanced Translate and Edit version, which enables the crowd member to edit both the source text and the target text, and score translation candidates within the group.

Feedback was positive, newly translated segments got high quality reviews, and the segments generated can also be leveraged in other translation workflows. One unexpected benefit of the tool: it is being used successfully to post-edit MT output and generate “clean” segments for future leverage.

Although it is still fairly basic, there are plans to add word processing, TMX and multi-format functionality to the translation tool. The entire ‘translation portlet’ is also being pitched as a standard feature on company internal portals, making translation a more natural part of the product and information experience.

Adobe, an early user of social translation, believes that its community drives product adoption. Communities cannot be dictated to: they are naturally self-organizing. The global technical expert community generates vast quantities of information, and multiple locales select want they want for rapid translation. Community leaders – not the company - choose and qualify translators and decide on how the community wishes to operate.

The Adobe community in China, for example, is a highly autonomous crowd of 70,000 members who select what they translate in terms of white papers, video subtitles, and articles. Partly for Internet access reasons, Adobe has found it hard to get feedback on what they need to do their job better. The community operates with its own logic.

Adobe therefore sees its task as providing tools to help these highly decentralized communities, and is partnering with Lingotek to develop a collaboration platform and a number of relevant tools. In 2008, it ran a small pilot of 21 translators for 125,000 words to identify needs, and discovered that the decentralized community model works, and that it is very much the communities and their leaders who run the show.

The platform, provided by Lingotek, featured a dashboard, and a translation interface, and there are plans to add more tools to meet demand for usability feedback, APIs and chat. Latency issues in accessing the US site also need to be overcome. Future moves include integrating the tool more tightly with content repositories and source document creators (echoing Cisco’s experience), engaging and training communities in Latin America and Spain, and stepping up its efforts to relate better to its Chinese communities.

Ant’s Eyeview, in the person of Sean Driscoll, wrapped up the ‘social’ session with a powerful plea to companies to embrace the user-centric model of localization beyond the corporate website. It’s not about “translation,” he said, “it’s about brand!”

While priorities in the mainstream model focus around website processes and flows, the new model is centered on an “emotion delivery vehicle” (think Facebook and Twitter) that brands use to find out how people experience their products.

Fragmented, decentralized content from multiple channels means that the “write it once and turn the crank” method will no longer be effective. Local markets will develop their own, community-sourced content, based on trusting other people rather relying on messages beamed out from corporate marketing.

The social shift to communities of passion impacts the translation industry. The company-driven model will have to give way to an organic approach. Management, control and incremental change will be overtaken by the organic processes of joining, engagement, low control and step change.

Whatever the activity, center of interest or community – call it the container - there is always just 1% of deeply engaged participants and 9% ‘editors’, while the remaining 90% form an audience of lurkers. There are four steps for successfully engaging with these containers and their disparate participants:

1. Decide on a clear strategy and objectives: Are you trying to satisfy customers, or gain loyalty or get feedback on your product or innovate or grow more efficient?
2. Listen and organize: Find out where the dialogs are happening, categorize and inventory the players.
3. Identify and engage with the creators, the critics and the connectors, and reward and recognize them appropriately.
4. Measure what you have achieved in terms of awareness, perception, engagement, and answers.

OTHER ARTICLES ON TAUS USER CONFERENCE 2009