Last week, the Association for Machine Translation in the Americas (AMTA), the European Association for Machine Translation (EAMT), and SDL announced the results of their joint 2009 survey on machine translation (MT). The findings confirmed what we found in our August 2009 report on the business case for machine translation — more consideration and use of MT, rising interest among language service providers (LSPs), and increasing awareness of the value of post-processed (aka “post-edited”) MT.

This is the second such MT survey run by these organizations, the first in late 2008, thus providing some data for ongoing comparisons. This year’s 228 respondents hailed from North America (57%), Europe (27%), and Asia (10%), and included Adobe, Agfa, Hewlett-Packard, Honda, Intel, and Kellogg’s. Their views echoed what Common Sense Advisory found in our interviews of MT users last summer:

• Acceptance. Twenty-eight percent of the AMTA/EAMTA/SDL study’s respondents are either using or plan to use MT. Common Sense Advisory’s research found increasing consideration and usage among organizations on both the demand side (corporate, government, and non-governmental organization) and supply side (LSPs).
• Drivers. Respondents look to MT for time and cost reductions. As we wrote last year, “There is too much content, too little money, and arguably not enough qualified people to translate every word into every commercially viable language.” Increasing translation volume, shortening turnaround time, decreasing cost, and improving information accessibility inspire most applications.
• Quality. Seventy-six percent of the survey’s respondents still worry about output quality. However, 25 percent said they would be happy with gisted output versus 60 percent who prefer “high quality, as though it had been written by humans.” We have found that even organizations with advanced implementations struggle with achieving quality. They employ increased more business rules or bilingual documents, multilingual assets like translation memory and terminology management, and human beings to improve their results.
• Post-editing. MT output that has been post-processed by humans was favored by 57 percent of the survey’s respondents. Our research confirms this. We found that such output should be a “no-brainer” decision for most companies in that it can easily deliver human-quality output at an appreciable discount in cost and faster turnaround time.
• Information security. Without naming names, the surveyors asked whether respondents would use “a publicly available, internet-based service to translate corporate documents.” Thirty-seven percent of respondents said “no” to using solutions such as Google Translate (or SDL’s own freetranslation.com?) for business-critical documents and intellectual property. Common Sense Advisory’s interviewees have expressed concerns about the possibility of their data seeping into the general pool of single-tenant free online MT sites.
• Language mix. More than 60 percent of the survey respondents targeted French, Italian, German, and  Spanish (FIGS) in their MT plans, followed by Asia languages (50%) and Portuguese (42%). Because our research shows that most organizations tend to increase language coverage over time, MT for FIGS could free up funds to translate second- and third-tier languages, an increasing mandate for global brands as they face regional and local brands (see “The Top 40 Global Online Brands,” Nov09).

Without a doubt, MT will help organizations deal with the burgeoning volumes of content that they face. As our recent research on trends in machine translation demonstrates, MT developers such as AppTek, Asia Online, Language Weaver, PROMT, and SYSTRAN are lessening the complexity of set-up and configuration, measurement and improvement of quality, and integration with other corporate systems (see “Developers Work to Broaden Use of Machine Translation,” Feb10).