Language Technologies in e-Commerce
Homework #2:  Answers

A. Internationalization

1. (5 points)
The .de version of Google has three radio buttons instead of two, allowing
the search to be restricted to just German pages as well as searching all
pages or all pages in German.

2. (5 points)
"Slicno strane".  Note that this is not actually true Serbian because it
is not written in Cyrillic (that difference of alphabets being the main
distinction between Serbian and Croatian, which otherwise are as similar
as British and American English).

3. (5 points)
"Zeese-a seerch terms hefe-a bee heeghlighted"

4. (5 points)
Reasons for country-specific sites:
  a. local staff that understands the culture
  b. local staff can be better at getting local advertising
  c. geographic proximity for better response
  d. legal -- different laws in different countries
  e. potentially less cluttered and thus easier to navigate than an
	internationalized site
  f. non-US users may be more likely to go to their own country's domain first

(Any two of the reasons above for full credit, any one for 3 points)

B. MT on the Web

5. (5 points)

InterTran offers the most distinct languages.

6. (5 points)

InterTran shows alternate translations for words when the user places the
mouse on special markers in the output.  For text (rather than URL)
translation, it offers rearrangement and selection among alternative
translations to improve the output.  There is also a link to add
new translations to a private dictionary. [full credit]

WorldLingo allows the user to specify a subject area to help improve the
translation with domain-specific data.  [3 points]

7. (50 points, 5 for rating each of the ten supported server/URL pairs)

There is a significant element of subjectivity in rating translation
services, but BabelFish and FreeTranslation are generally the best of
the four, while InterTran definitely has the lowest quality.

Some time during the week between assignment and due date, something
changed for InterTran, and it was no longer able to produce a translation
for the Spanish URL, instead just hanging until the browser generates
an error.

WorldLingo has a bug, and translates the wrong page for the Spanish
URL (two points deducted if there is no mention either here or for Q8
that different text was translated).


8. (8 points)

All of the translators do a good to excellent job of preserving the
formatting.  InterTran adds the alternate-translation icons after
nearly every word, which is distracting when reading the text.  All of
the translators suffer occasionally when words are reordered around
HTML markup, leading to extra or missing words in link, bolded phrases,
etc.

9. (6 points)

MT is definitely usable for information-gathering, as one can understand
the general content of a translated page, though usually fine details are
lost.  The near-instantaneous translation time makes it practical to use
when it is not possible to wait for a human translation.


10. (6 points)

Due to grammar and translation errors, none of these systems as yet
has sufficient quality to create pages that a business would want to
use directly for its e-commerce site, although some other less-general
translators are coming close.  The better translators MIGHT be
acceptable for publishing personal pages in multiple languages,
depending on the complexity of the page and the user's threshold for
errors, and can definitely improve a human translator's productivity
by providing an initial translation that can be fixed up.