26
Mar
Benjamin B. Sargent 26 March 2010
Filed under (Language Industry, Translation Technologies)
2 pepper rating

Last week, 100 attendees — largely owners and operators of language service providers (LSPs) — gathered in Phoenix, Arizona for the 11th annual ATA-TCD conference (that stands for American Translators Association, Translation Company Division).

Highlights of the show included presentations on sustaining high growth, trends in machine translation, technology planning, and quality standards, along with a look at how to protect language workers from harm and retaliation by advocacy organization Red T. Common Sense Advisory participated in a formal debate on how LSPs should approach technology strategy. The bookend talks served to illustrate the topics and tenor of the meeting.

The opening keynote by Jost Zetzsche identified four perceived threats facing translation companies:

  1. First, while crowdsourcing remains a hot topic, the speaker noted that crowdsourcing only works in specific areas and its overall impact on the industry will be limited – despite several high-profile cases. He suggests that company owners should recognize the benefits buyers seek from crowdsourcing and develop services with those characteristics.
  2. About the second threat, machine translation (MT), Zetzsche made the observation that generic (including free) translation will always be sub-par but that customized engines in association with memories in fact represent a useful technology appropriate for many applications.
  3. Thirdly, he observed that only larger, sophisticated organizations can take advantage of data sharing today. Moving forward, smaller agencies and even freelancers need to learn how to unlock value from shared assets rather than be mere bystanders.
  4. The last threat he discussed was the notion that “ecosystems” are being created by Lionbridge and SDL through technology and partnership programs, the inference being that smaller agencies, by choice or by accident, are either on the inside or on the outside of those respective business networks.

The message behind all of Zetzsche’s points was that each threat also represents an opportunity for service innovation. The closing session was by Kirti Vashi of Asia Online furthered this argument, laying out specific ways that LSPs should be embracing MT rather than resisting it. Specifically, he encouraged attendees to look at MT + human translation (HT) rather than MT vs. HT and to visualize the “highly adaptive human-driven process for continuous output quality improvement in statistical MT engines” as one of the roles an LSP plays in the supply chain — not just for now, but in perpetuity as languages and products evolve. The new markets for translation that emerge as the pricing waterline recedes will flood the industry with new revenue opportunities. But, service innovation is required and organizational competence must be built.

Here is our takeaway from the MT discussion — there is a continuum between static, high-value content to ephemeral, low-value content. Historically, HT operates at the high-end and “zero translation” (ZT) is found at the low end. What’s happened in recent years is that a range of MT options finally crossed the chasm from dis-utility (not useful) to utility (good enough).

  • On the low end of the market, free tools like Google Translate and Bing Translator now produce reliable “gist” translations in a dozen or more common languages — to and from English — with many more pairs moving quickly up the quality curve toward utility (this is an important revenue source for information publishers).
  • In mid-range applications, customized engines are providing utile raw output for commercial purposes such as customer support articles. In these applications, MT users like Microsoft process vast volumes of information relevant to customers where HT is not viable due to time and cost realities.
  • Finally, on the high-end of the market, MT is being added to human translation workflows in order to boost productivity. Where quality and durability of published information is a requirement, some translation providers in some situations have demonstrated the utility of post-processed or post-edited MT, although we still hear from others that report making investments without a positive return.

Last week at the ATA-TCD conference, many of the presentations and discussions returned to MT — as in previous years. The difference this time was in how MT is regarded. No longer a vague and looming threat in the future, translation companies now grapple with MT as current reality. While questions remain about the long term consequences of MT adoption in the industry by enterprise translation buyers and by end-consumers, the initial tidal wave has already hit the beach.