2012-01-03 22:03, Dr J R Stockton wrote:
> In general, a Google Translate is good enough for a
> monoglot reader to tell whether the site is of interest, and perhaps to
> obtain a general rough understanding.
Perhaps, but what is the added value of a site-specific translation
widget? A person who does not understand the language of your page will
seldom stumble across it. If he finds it with a search engine because he
used suitable proper names or international words, then the odds are
that Google suggested a translation possibility to him. How will the
poor user know whether the translation widget on the page is something
better? Besides, by leaving it to translation systems to advertise
themselves, your site won't get the moral responsibility for whatever
absurd and wrong translations might be generated.
As I previously wrote, in special cases, for well understood purposes,
translation widgets can have real added value. As a paradigmatic (i.e.,
imitated) "standard" feature of pages, they are comparable to ideas of
having interfaces to search engines as part of a normal web site or to
"print this page" or "bookmark this page widget".
> That is better than nothing,
Wrong translations are often worse than no translation, especially if
they look good. This is part of the problem with Google Translate: it
uses heuristics and statistical methods and often processes a longish
phrase as a unit, often resulting in syntactically good-looking text.
But it may get the content completely wrong; I've seen it translate a
word for Monday to "Sunday" (probably due to errors in its analysis of
texts).
> My home page has such a button, and my site can therefore be read in,
> for example, something vaguely resembling Finnish.
I would be rather surprised to find a Finn who has interest in and
chances of understanding the contents of your pages and who understands
the "Finnish" generated by Google Translate better than your English
text. Elder people (over 50 or 60) may have read German or Latin as
first foreign language and may have lacked the opportunity to learn
English at school or later, but I don't think many of them would be
interested in MS-DOS Batch files or JavaScript programming. Even if
there were such a person, why would he understand the word "Translate"
or the words "Select Language"?
> Its idea of the Finnish for Kepler's Laws translates back to something
> at least recognisable;
Back-translation using the same automatic translator proves very little
especially in the positive direction, since translation in one direction
probably shares quite a lot of logic, code, and vocabulary with the
opposite translation.
The translation, once got to it after using the translation widget on
your main page (which seemed to make it impossible to follow links
without opening them in a new window or new tab), is total nonsense. A
majority of the words are sensible translations of the English words,
but the sentences just don't make sense - only the very first words "the
orbits are ellipses" have been translated intelligibly.
To be honest, the Swedish translation is rather understandable, with
just a few grammar errors and with some obscurity caused by pronouns (so
typical of machine translation: how could the software know what "its"
refers to?).
> And, of course, unless one has a full team of translators at one's
> elbow, only machine translation can track changes in near-real-time.
The alternative to machine translation widgets is not the construction
and maintenance of a hundred different language versions. Rather,
leaving possible translation outside the scope of the site - users can
use it if they like, using a translator of their choice. You may help
this somewhat by keeping the language as simple as possible (but not
simpler), using spelling checkers (spelling errors unnoticeable to human
readers may thoroughly upset machine translation), etc. - and by
following the rule that as far as feasible, the text content of each
HTML element, such as link text, should constitute a separately
understandable and translatable expression.
--
Yucca,
http://www.cs.tut.fi/~jkorpela/