 |
|
|
|
Admerix is a pioneering provider of post-machine translation editing.
• Experience in reviewing and editing MT output
• Intimate knowledge of specific issues that come up with MT translation post processing
• Domain-specific subject specialists available
The machine translation (MT) revolution
The machine translation era is coming quickly and the signs are that the process is evolving at a much faster pace than has ever been predicted.
The main players are Google, Language Weaver, Microsoft, Prompt, and Systran. The hot topic for debate is who is currently delivering the most accurate translations across the range of texts likely suitable for the process.
Actually, that topic in the overall scheme of things is irrelevant. The fact is that the technology is evolving and will continue to evolve at an ever increasing speed. The end result of this is that within the very foreseeable future, machine translation is going to be a fact of life and a very useful tool.
In a much shorter time than many people expect, localization buyers will be demanding that their costs and timelines be reduced through the implementation of MT technology.
Applications for MT technology
There are the obvious applications for machine translation such as
• large websites with volumes of information that need to be disseminated to a much wider international readership
• time-critical press releases which need fast turnaround and distribution
• chat rooms and forums with participants speaking different languages
For these applications of MT, the actual quality of the translation work is not critical as long as it is reasonably readable and accurate. The intention and basic meaning of the text is the critical factor.
Foreign language knowledge bases
MT will generate its own market in many ways that have not even been thoroughly explored yet. There is a rapidly expanding amount of technical knowledge, research data, legal case documents, and patent documents that are being posted online or otherwise made available to businesses with the specific interests.
Access to this huge amount of information is currently limited because of the specific languages in which the information is recorded. Such texts, if made searchable in English (and other languages as well), will open all the data up for the use and benefit of many more people. To have all of these fully translated in the normal way just to allow the searches to be conducted is not only impractical, but out of the question on a cost and time basis.
|
|
|
|
Machine translation offers a practical opportunity to have such materials basically translated so that they can be stored in a searchable system. When relevant data is isolated, only those specific documents can then be given over to the full human translation and checking process, thus saving a great amount of time and cost.
Technical documentation is also a candidate for MT where domain-specific memories can be created. Since technical specs have specialized terminologies, domain databases can be constructed to narrowly apply to company or subject-specific projects.
Options when considering an MT solution
In order to get the best possible use for MT output, there are several possible methods:
1. No editing or human intervention.
2. Minimal fast proofread of translation only for general readability and sensibility.
3. Editing of output translations against source text for accuracy and general readability and sensibility.
4. Careful professional editing by highly qualified linguists for ultimate translation quality and accuracy. (The holy grail for all developers of MT systems is to reach a translation output good enough to qualify for this process).
Of the above processes, only the first one does not require a talented linguist. The cost of each option is, of course, in line with the desired outcome, but no matter how it is approached, the aim is that it will be faster and cheaper than conventional localization processes.
|
