I did a similar project BabelCode for over a year. It died of impracticality this September. Since I'm "out of competition" now, I feel free to tell the deadly disadvantages of such systems:

(1) Only one-way communication. You may use a controlled authoring and translation system to write to a foreign audience, but you can't read information in an original foreign language with it. It's almost impossible to make an American to talk to someone in the rest of the world using a controlled translation system. Without learning English, an ignorant nation will continue to keep ignorant. So this disadvantage limits the use of the system to only multilingual document generation for the few big companies.

(2) Learning to express a natural language sentence in a formal language system is no less difficult than learning to fluently develop applications using a certain software development kit (SDK). Even if you let the user input formal language expressions directly, either you have to preset many rules (and "super-rules" that define solutions to rule conflicts) to capture all possible ways in which the user may construct formal expressions for a certain phrase pattern, or you have to force the user to construct formal expressions in only one or two possibilities acceptable by the system. The latter is more practical for the system but moves the burden to the user; that's why a Win32 programmer has to consult MSDN and search the Usenet for relevant discussions (and sample code therein) before he could reliably develop for a certain application field (e.g. Multimedia, Networking). Given the learning curve and complexity of writing for a certain topic using a controlled translation system, there is no wonder how less efficient it would be compared to learning the target foreign language (e.g. English) instead.

The above two disadavantages makes controlled translation systems impractical to be adopted by general users. However, my past research project did spawn an invention named LingoX.

LingoX is a foreign language writing aid inspired from the controlled language authoring process and, more indirectly, from the "IntelliSense" feature seen in Microsoft Visual Studio.

When the user types in a foreign language anywhere in Windows, the program intercepts keyboard input and display under the blinking cursor the "usage syntax" information for the last word/phrase he has typed. For example, if the user types: I am searching, then the program prompts:

v. search [n. Site] [for n. Target]
n. search [for n. Target]

This way, the user gets fully informed of how the word can extend the current sentence. For each "usage syntax", the "syntactic requirement" (e.g. "n."), "semantic requirement" (e.g. "Site") and "morphological requirement" (e.g. "adj. dedicated [to doing sth.]" tells the user not to use the pragmatically incorrect construction "dedicated to do sth.") for each "argument" of the headword.

The program can pre-emptively prevents many syntactic, semantic, morphological and pragmatic errors that can't be detected by tradtional spell checkers and grammar checkers with certainty. Fully and explicitly prompting the "usage formulas" also avoid frequent lookups in an online dictionary and the time-cosuming "manual formula extraction" from example sentences provided by a dictionary. Of course it also prevents the user from boldly "inventing" his own incorrect usage by guessing. It also exhibits the "beauty of mechanics" of a foreign language, incenting the user to explore more new words and practise more writing. It also ....

The program will be available on www.lingox.com in the near future.

Best Regards,
Yao Ziyuan
Second-Year Computer Science Student of Fudan Univ.

From: "Donald Z. Osborn" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Subject: [Mt-list] Fwd: Universal Networking Language (UNL), other Interlinguas & their Applications
Date: Sat, 27 Nov 2004 19:19:09 -0600


The following, seen on the linguist list, may be of interest (UNLs & MT is a
topic).


Don Osborn
Bisharat.net


Date: 21-Nov-2004 From: Alexander Gelbukh <[EMAIL PROTECTED]> Subject: Universal Networking Language (UNL), other Interlinguas and their Applications


Full Title: Universal Networking Language (UNL), other Interlinguas and their
Applications


 Date: 16-Feb-2005 - 16-Feb-2005
 Location: Mexico City, Mexico
 Contact Person: Alexander Gelbukh
 Meeting Email: [EMAIL PROTECTED]
 Web Site: http://www.cicling.org/2005/UNL.htm

Linguistic Field(s): Applied Linguistics; Computational Linguistics; Language
Description; Lexicography; Semantics; Syntax; Text/Corpus Linguistics;
Translation


 Call Deadline: 30-Nov-2004

 Meeting Description:

This workshop will be devoted to the presentation of various approaches to the
development and use of interlinguas. Interlinguas are primarily developed for
machine translation, but it is obviously not the only application in which a
meaning representation formalism, (relatively) neutral with respect to
language, can be used. Other applications include information retrieval and
dissemination, summarization, knowledge mining, and multilingual generation.


Among the existing interlinguas, special attention will be given to UNL
(Universal Networking Language). This language is the basis of a large
international effort initiated by the United Nations University and aimed at
multilingual communication on the Internet. We hope that the groups engaged in
UNL activities will present the current state of affairs.


The workshop seeks original papers on all aspects of interlinguas. Topics of
interest include, but are not limited to, classical issues and novel
applications, such as:


* Interlingua-based MT systems and multilingual generation.
* Issues related to novel applications of interlinguas: information
dissemination, retrieval and extraction based on interlingual representations,
creation of multilingual documents and their integration into Internet-based
dissemination systems.
* Interlingual representations and interlingual lexicons as linguistic resources
that can be integrated into NLP systems, and, conversely, existing resources
which can be incorporated into interlingual systems.
* State of the art of the UNL modules for various languages.
* Evaluation of interlingual representations.
* Multilingual applications and services.
* Internet-based multilingual information dissemination.
* Multilingual information retrieval and information extraction
* Multilingual documents processing.


 PROGRAM COMMITTEE

 Pushpak Bhattacharyya
 Igor Boguslavsky (chair)
 Christian Boitet
 Jes�s Carde�osa
 Mike Dillinger
 Alexander Gelbukh
 Irina Prodanof
 Virach Sornlertlamvanich

 ORGANIZERS

 Jes�s Carde�osa
 Alexander Gelbukh (chair)
 Edmundo Tovar

 PUBLICATION

All the accepted papers will be published in a special issue of the journal
Research on Computing Science, ISSN 1665-9899. This issue will be also supplied
with an ISBN.


 SUBMISSION

The format for submissions is the same as for CICLing-2005 conference
(http://www.CICLing.org), see Format instructions there (ignore the paper size
limit indicated there).


 Submissions are to be sent by email in pdf or rtf format to:
 Edmundo Tovar ([EMAIL PROTECTED]).

 AGENDA

The workshop will consist of two half-day sessions. For the first session, we
invite submissions related to all aspects of the interlingua development. The
second session will be devoted exclusively to UNL and will be of more practical
orientation. We will begin by identifying some linguistic phenomena which
caused difficulties for UNL encoding in various experiments and provoked
disagreement among the encoders. Then we will propose and discuss some ways of
overcoming these problems and, finally, we will have a practical encoding
exercise aiming at testing and solidifying solutions arrived at during the
discussion.


 IMPORTANT DATES

30th November, 2004: Deadline for submissions
15th December: Notification of acceptance
10th January, 2005: Firm deadline for final version submission (camera-ready)


 Tentatively 16th February: Workshop

 CONTACT AND MORE DETAIL

 See http://www.cicling.org/2005/UNL.htm
 or http://www.CICLing.org/contact.html




_______________________________________________ MT-List mailing list [EMAIL PROTECTED] http://www.computing.dcu.ie/mailman/listinfo/mt-list



_______________________________________________ MT-List mailing list [EMAIL PROTECTED] http://www.computing.dcu.ie/mailman/listinfo/mt-list

Reply via email to