Re: English to Japanese Machine Translation

Subject: Re: English to Japanese Machine Translation
From: David Neeley <dbneeley -at- gmail -dot- com>
To: "TECHWR-L" <techwr-l -at- lists -dot- techwr-l -dot- com>
Date: Mon, 1 Aug 2005 12:33:31 -0500


As it happens, I am involved in preparing a large proposal for a
machine translation program to translate many languages into
functional English.

It turns out that the task of translating other Indo-European-derived
languages to and from English is a major undertaking, but translating
to and from a language of other linguistic bases is even more
difficult.

Many Japanese technical documents contain essential bits in English,
as it happens, because Japanese is by its nature extremely
non-specific--depending largely upon context for meaning. (Japanese is
not quite as difficult in this regard as Chinese, as it happens).

Some years back, I did a good bit of consulting at the El Segundo
campus of Xerox--where I saw many technical documents originating in
Japan at their joint venture, Fuji Xerox. In them, even before
translation, liberally sprinkled through the documents were English
specifications.

In moving from English to Japanese, an MT program must be
sophisticated enough to deal with various ways of expressing things
depending upon the combination of characters selected. It can be a
daunting task.

Most MT programs today fall under the category referred to as
"transformers" with a few more elaborate ones falling into the
"linguistic knowledge" systems. Our research indicates that only small
and incremental improvements can be expected from these approaches,
and that they are far less good when translating between the
Sino-Tibetan languages (including Chinese and Japanese) and English.

That is why we are examining a different methodology entirely, based
upon an old concept that has not been practical until recently (an
Interlingua). However, that would be a major project over multiple
years--even though our target is only one-way translation, and even
then the initial objective is an 80% accuracy.

For a technical document, that would be insufficient, obviously.
However, for our project it would be a substantial beginning, as it
would highlight documents needing the specific attention of
native-level speakers of the target language.

In addition, I have a broad experience in using MT programs in
translating back and forth between English and Russian. With
experience, it becomes far easier to use MT in a meaningful basis.
However, to the casual user it can be extremely frustrating.

If you doubt this, try Babelfish to translate short passages between
English and a second language--then use the output of the translation
as the input to translate back to English. You will likely be rather
amused at the result. Babelfish uses Systran, by the way.
http://babelfish.altavista.com

David

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Now Shipping -- WebWorks ePublisher Pro for Word! Easily create online
Help. And online anything else. Redesigned interface with a new
project-based workflow. Try it today! http://www.webworks.com/techwr-l


---
You are currently subscribed to techwr-l as:
archiver -at- techwr-l -dot- com
To unsubscribe send a blank email to leave-techwr-l-obscured -at- lists -dot- techwr-l -dot- com
Send administrative questions to lisa -at- techwr-l -dot- com -dot- Visit
http://www.techwr-l.com/techwhirl/ for more resources and info.



References:
Re: English to Japanese Machine Translation: From: Bill Swallow
RE: English to Japanese Machine Translation: From: James Jones

Previous by Author: Re: QUERY? Comment?
Next by Author: Re: Certification--a new concept. Discuss
Previous by Thread: RE: English to Japanese Machine Translation
Next by Thread: Re: The one question I hate the most


What this post helpful? Share it with friends and colleagues:


Sponsored Ads