Machine Translation
usually refers to the use of computers and software for the translation of a
text/message from one language to another.
In its beginnings in
the 1950s, machine translation relied on the substitution of words in the
target language. But, since this led to
strange constructions and failed to look like a natural language, machine
translation relied on corpus and statistics.
One of the main problems is the differences in language typologies. Languages from two different families present
more difficulties to machine translation attempts.
However, the use of substitutions
is still in place. One way to make this
substitution less risky is to use sensors or identifiers of the lexical field
in question. If the text talks about a kitchen
description, for example, the word ‘sink’ is translated appropriately into
Arabic, rather than using a word which refers to drowning in the sea/water
pool. Of course, a professional Arabic
translator won’t need a lexical categorization in order to identify which term
is meant!
By relying on human
intervention, machine translation can be made more accurate and reliable. The identification of the word class of some
problematic words makes it easier for the output to be less confusing. Some Arabic names are translated into nouns
and adjectives, and this makes the output more confusing.
Moving from being a
mere illusion to some practical achievements has made Machine Translation more feasible
and acceptable for users. People claim
that the product of Machine Translation can be safely used in some domains. But, if the frequency of error is important
and if the possibility that errors can happen, the reliability level remains
very limited, especially in scientific, legal and technical domains.
However, the
achievement of a totally professional output seems more difficult than
realistic. It remains reasonable to
think of MT as a tool any users can apply in order to get the gist of a text,
especially when the source and target languages are from the same language
family. Accuracy and professionalism
require going beyond the mere replacement of words to the artistic, stylistic
and skillful linguistic formulation of a message in another language. Style and language-specific nuances are not
easily achieved by human translators, not to mention the limitations of
computer software. Hence, these limitations
remain inherent in all MT applications. This
applies to literary texts, normal interactions in everyday language as well as
to texts in humanities and social sciences.
Although automated translation has shown promising results, prospects
seem very limited since the integration of an experienced smart individual with
a native communicative competence into software is far from being achievable.
One more issue is that human translation goes
through a quality assurance process. Every
translation needs double checking and an external proofreading in order to be
professional and reliable. However, MT
cannot meet these quality requirements. In fact, sending machine translations to
proofreaders to check, apart from aiming at reduce the cost at the expense of
the translator’s profit, reveals that translation is an human art, and not an
aspect of automation.
It is worth
mentioning here that CAT tools are different from machine translation
(MT). Computer Aided Translation aim at
helping human translators make perfect translations with more output and
consistent terminology. Professional translators
need them to improve their daily output and employ the same terms in the whole
document.
Almiaad Lingua for Translation & Language
Services, a global provider of professional translation with a local perception!
Isn't this an attempt to save translators from early retirement?
ReplyDelete