Automatic Translation Percolates
April 15, 2010
SDL is a company that provides global information management to organizations worldwide. The firm is active in automated translation, Web content management, structured content technologies, and eCommerce. I learned that SDL reported some of the findings from the firm’s study into machine translation. The factoids in this write up come from “SDL Reveals Results of its Automated Translation Survey”:
- A surprising 28 percent of those in the sample of 228 people are using or plan to use automated translation. (My anecdotal information about machine translation was that its use was well over 50 percent. Part of this perception comes from comments made about the uptake of Google’s free translation service and its automated method in the Chrome browser.)
- The reason the survey sample does not use machine translation is related to quality. (The use of human translators does deliver better handling of slang, particularly in casual communications and some types of informal writing. But machine translation has, based on our tests at ArnoldIT.com, works quite well on scientific, technical, and medical information and certain types of formal business writing; for example, a proposal that conforms to a specific set of technical guidelines set forth in a statement of work.)
- More than half of the respondents wanted to hook machine translation to human translators in order to improve quality.
The article said:
SDL invested in machine translation in 2001 and launched its Knowledge-based Translation System in 2004. SDL publishes over 7 billion words of content through its automated translation systems every year. A 300 strong team of computational linguists, project managers and post-editors has been human post-editing machine translation for global clients for over 6 years. SDL sees the success of machine translation as being through its integration in the translation process. SDL integrates machine translation technology with consulting services, desktop translation memory and enterprise translation management systems in hosted, on-premise or SaaS environments to suit the needs of global business.
Beyond Search believes that the volume on content to translated from source to target language makes human translation more problematic. In one police department, the organization has more translators than a major Federal agency. The reason is that certain languages lack a sufficient number of translators for that language. Google’s free system relies on humans, but the company invites a person to submit a better translation that the one Google produces.
Costs of human translation continue to rise. Interesting situation with content volume increasing, humans becomes more expensive, and free services available.
Stephen E Arnold, April 15, 2010
Unsponsored post.
Comments
2 Responses to “Automatic Translation Percolates”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
some of the facts here does not seem to be true:
1. “Costs of human translation continue to rise” – Not true. Costs of human translation is the lowest it has ever been (see http://www.tomedes.com as an example).
2. “machine translation has, based on our tests at ArnoldIT.com, works quite well on scientific, technical, and medical information and certain types of formal business writing” – Simply Not True
@Jim: 2. “machine translation has, based on our tests at ArnoldIT.com, works quite well on scientific, technical, and medical information and certain types of formal business writing” – Simply Not True
Jim, how do you know whether Stephen’s assertion is true or false? For some types of text, machine translation IS satisfactory.
For an evaluation of the quality of free machine translation from Google, see http://www.international-english.co.uk/mt-evaluation.html. Professional translators evaluated the texts. The translations from English to Spanish and from English to Norwegian are satisfactory. The translation from English to Welsh is not good.