Representatives from Globalink, Alis and SYSTRAN use such phrases as 'not perfect' and 'approximate' when describing the quality of translations, with the caveat that sentences submitted for translation should be simple, grammatically accurate and idiom-free.
'The progress on machine translation is moving at Moore's Law — every 18 months
it's twice as good,' said Vin Crosbie, a Web industry analyst in Greenwich,
Conn. 'It's not perfect, but some [non-English speaking] people don't realize
I'm using translation software.'
With these translations, syntax and word usage suffer, because dictionary-driven databases can't decipher between homonyms — for example, 'light' (as in the sun or light bulb) and 'light' (the opposite of heavy).
Still, human translation would cost between $50 and $60 per Web page, or about 20 cents per word, SYSTRAN's Sabatakakis said.
While this may be appropriate for static 'corporate information' pages, the machine translations are free on the Web, and often less than $100 for software, depending on the number of translated languages and special features."
4.3. Computer-Assisted Translation
Within the World Health Organization (WHO), Geneva, Switzerland, the Computer-assisted Translation and Terminology (Unit (CTT) is assessing technical options for using computer-assisted translation (CAT) systems based on "translation memory". With such systems, translators have immediate access to previous translations of portions of the text before them. These reminders of previous translations can be accepted, rejected or modified, and the final choice is added to the memory, thus enriching it for future reference. By archiving daily output, the translator would soon have access to an enormous "memory" of ready-made solutions for a considerable number of translation problems. Several projects are currently under way in such areas as electronic document archiving and retrieval, bilingual/multilingual text alignment, computer-assisted translation, translation memory and terminology database management, and speech recognition.
Contrary to the imminent outbreak of the universal translation machine announced some 50 years ago, the machine translation systems don't yet produce good quality translations. Why not? Pierre Isabelle and Patrick Andries, from the Laboratoire de recherche appliquée en linguistique informatique (RALI) (Laboratory for Applied Research in Computational Linguistics) in Montreal, Quebec, explain this failure in La traduction automatique, 50 ans après (Machine translation, 50 years later), an article published in the Dossiers of the daily cybermagazine Multimédium:
"The ultimate goal of building a machine capable of competing with a human translator remains elusive due to the slow progress of the research. […] Recent research, based on large collections of texts called corpora - using either statistical or analogical methods - promise to reduce the quantity of manual work required to build a MT [machine translation] system, but it is less sure than they can promise a substantial improvement in the quality of machine translation. […] the use of MT will be more or less restricted to information assimilation tasks or tasks of distribution of texts belonging to restricted sub-languages."
According to Yehochua Bar-Hillel's ideas expressed in The State of Machine Translation, an article published in 1951, Pierre Isabelle and Patrick Andries define three MT implementation strategies: 1) a tool of information assimilation to scan multilingual information and supply rough translation, 2) situations of "restricted language" such as the METEO system which, since 1977, has been translating the weather forecasts of the Canadian Ministry of Environment, 3) the human being/machine coupling before, during and after the MT process, which is not inevitably economical compared to traditional translation.