Certainly all content does not need to be translated into multiple languages. For example, the market for translating Italian classified automobile listings into English is relatively small. However, the case could be made that exotic car dealers in the UK benefit from reading those listings in English, particularly if they are looking for a rare find for their clientele.The question to translate or not to translate is usually a function of the following factors:
- Cost
- Lead time: can the content be translated before it perishes?
- Comprehension level: does the translation require a specialist (e.g. Medical or Legal translations) or can it be simply understandable?
- Market return: what is the value of the translated information?
Up until 1992, the exotic car dealer in our example had only one option to turn too for translating his listing into English: hiring a professional Translator. Today, the paradigm has shifted: with the click of a mouse, our car dealer can instantaneously obtain a usable machine translated version of the listing in English. This is not to say that the MT translation does not have tradeoffs, which are highlighted in the following chart.
| |
Professional Translation* |
Machine Translation |
| Translation Rate |
A translator can translate 2,500 words per day, on average; a translation team can be scaled to handle substantially larger volumes |
Commercial systems handle 300 to 500 words per second |
Quality
Output |
Depends on the linguistic expertise of translator(s). Perfect translation quality to this day is only achievable through professionals. |
Comprehension level ("Gist" quality) or better with out-of-the box software; quality improves with Linguistic Training of the system |
| Cost |
$0.15 to $0.45/word, depending on language and type of document |
Fractions of a penny per word |
Technical
Scalability |
Graphic design and re-engineering (e.g. l10n or i18n) required for certain applications |
MT can be integrated directly into application (s) in most instances |
| Formats |
All formats with the aid of a designer |
Most common formats : HTML, TXT, Notes, Otext, DOC, RTF, etc. |
| Languages |
All languages |
Extensive language coverage |
| Changes |
Each edit in the source requires retranslation |
Edits also require re-translation (a click of a button) |
| * The use of Translation Memory or Globalization Management Systems (GMS) can automate certain tasks. |
Table 1: Key differences in Professional versus Machine translation
The tradeoffs highlighted in Table 1 should be used by decision-makers to determine whether or not their content is MT-worthy. All too often, vital information that does not get translated could have been rendered usable in other languages via MT or, yet again, content that was professionally translated at a high cost may have been translated at a lower cost with MT. MT and Professional translation can also be used in combination, in a layered approach. IBM uses both quite successfully on its software support site. On IBM’s site it is financially unfeasible for all FAQs, postings and other content to be professionally translated. Key benefits of this approach are:
-
Machine translated content does not need to be stored- it gets generated on demand
-
The latest source content always gets translated and
-
It is self-manageable.
Certainly, most content requires a level of quality that MT on its own cannot offer. Nonetheless, many businessmen get by with just the most basic knowledge of a language while overseas. And MT is light years beyond High School French. Although new uses for MT are constantly surfacing, it is particularly well suited to those areas where, due to cost, time, technical unfeasibility or volume, professional translation is not an option. A non-exhaustive list of those applications is as follows:
- Instant exchanges: Email and chat collaboration (e.g. Lotus Sametime)
- Knowledge sharing: Portal, Content Management Systems (CMS), Search and retrieval
- Support: Publication of instructions, FAQs, help files
- Intelligence mining
- Public service: Job postings, electoral, health care, unemployment and disability information, etc.
- Any area where language prevents the natural flow of business (some Web applications, CRM system…)
The bottom line is thus: the addressable market for translatable content is very large while only a fraction of that content currently gets translated today. The stage seems set for MT. But is it really? The main factor holding back further investment in MT solutions is quality. |
