Stefán Briem

 English homepage                     Dansk       Esperanto  Íslenska

What is machine translation?

Machine translation of written text from one language to another means using purposeful computer programs, that input a text in a certain language, the source language, and deliver its content in an equivalent text in another language, the target language.

If the program works independently, without human interference from the time the program starts until it delivers a result, the process is called a (fully) automatic translation. Usually, however, a better quality is achieved from a translation with a combined effort of the computer and the human mind. It can e.g. happen in that way, that the computer places a question to the user of the computer, when a matter of uncertainty arises, and waits for an answer, resuming processing when it has received the answer. That is called a semiautomatic translation. Another method is, that a computer delivers a raw translation, whereafter a human translator takes over, corrects errors and improves the syntax or language. And still another method consists in pre-editing the text before it undergoes machine translation in order that the translation program better can cope with the text.

Machine translation is the main field of language technology and is connected to most of its other fields. Research and development in machine translation started about five decades ago, but the progress has proved to be much slower than was expected at the beginning. The most difficult factor in machine translation is the semantic analysis, i.e. making a program “understand” the meaning of the text to be translated. The quality of machine translation strongly depends on the success of coping with this factor. Computers will never fully replace human translators, but they can already speed up and facilitate the translators’ work considerably, at least when the text is limited to specific subject fields or with other limitations put on the input text. In Iceland, until now, little effort has been put into the field of machine translation, except by amateurs in their leisure time.

In which ways can machine translation be useful?

The main advantages of machine translation compared to the work of human translators are, that a computer works much faster than a person and that by machine translation it is easier to keep consistency in the translation.

Sometimes it can be sufficient to get an unprecise translation, that reveals what the text is about without everything being translated correctly. And at the same time it can be more important to get the result without delay than to get a translation of good quality. An example of translations of this kind are the translations offered free of charge for a few languages by the AltaVista search engine.

The processing speed in machine translation can be highly important, although it has to be followed up by human translators correcting the result and improving the syntax and language. The advantage of machine translation can then show up as well in less expenses due to less effort of human translators as in shorter total time for the translation than if performed exclusively by humans.

What is customized machine translation?

Sometimes it can be appropriate to customize machine translation for certain tasks. For example, it can be a question of preparing a special word list before a translation is started, where regard is taken to the subject field of the text and its style. With such a special word list, the weight on certain translations of ambiguous words or word combinations can be changed and notions can be added, when they are not found in the basic word list of the translation system to be applied. If the text contains a lot of repetitions, it comes into consideration to incorporate whole sentences or phrases in the special word list. Whether, or in which way, it is appropriate to customize the translation method, depends on the goals, circumstances and the type of text to be translated.

Which method is appropriate in each case?

In order to estimate which translation method is appropriate for a certain task, the discussion presented previously has to be kept in mind, and it has to be considered carefully if and then to which degree it can be appropriate to apply automatic methods.

Generally it can be stated, that machine methods can be appropriate:

  • for technical/informative texts (but much less for literary texts)
  • for texts containing a lot of repetitions
  • for texts where it is important to keep consistency
  • for texts with small vocabulary
  • when a task has to be completed quickly

  • Last changed 30 September 2000