Presentation in the seminar Applications of Computational Linguistics by Susanne Peters, 9 November 2001
[structure] [bibliography]

Applications of

Computational Linguistics

Machine Translation:


 
 


 
 
 
 

STRUCTURE OF THE PRESENTATION

1. SYSTRAN: an outline
+ examples of free real-time translations powered by SYSTRAN

2. SYSTRAN - Past and Present:
a brief history of SYSTRAN translation software

3. How does SYSTRAN work?
summary of the article by Peter Wheeler 'Systran', 1987

4. The Implementation of SYSTRAN at XEROX:
summary of the article by John Elliston 'Computer Aided Translation: A Business Viewpoint', 1979

5. The Implementation of SYSTRAN at the Nuclear Center of Karlsruhe (Kernforschungszentrum Karlsruhe)
summary of the article by F.W.A. Habermann 'Provision and Use of Raw Machine Translation', 1986

6. The Role of the Posteditor:
summaries of the articles by
Roy Green 'The MT Errors Which Cause Most Trouble To Posteditors', 1982;
Bernhard Lavorel 'Experience In English-French Post-Editing ', 1982;
Barbara Kostrewski and Peter Taylor 'Discussion on Session 3: The Posteditors' Experience', 1982

 
Babel Fish Translations

http://babelfish.altavista.com/

free real-time translations on the Internet powered by SYSTRAN
 
 
 
 
 

SL

SYSTRAN is the most widely used MT system to date. It is extremely successful, although it is of early, rather primitive design. This means that a first generation direct approach is typified in SYSTRAN, as opposed to a second generation interlingual approach and a third generation transfer approach. We have heard about this last week. In a direct approach a specific language pair is considered, and emphasis is put on a bilingual dictionary as knowledge source. SYSTRAN's success is largely the result of the great effort that has been put in over many years to incorporate linguistic knowledge into its system.
 
 


TL

SYSTRAN ist das allgemein verwendete maschinelle Übersetzungsystem bis jetzt. Es ist, obgleich es von früh ist, ziemlich ursprüngliches Design extrem erfolgreich. Dies heißt, daß eine erstes Erzeugung direkte Annäherung in SYSTRAN, im Vergleich mit einer zweites Erzeugung interlingualen Annäherung und einer drittes Erzeugung Übergangsannäherung verkörpert wird. Wir haben über diese letzte Woche gehört. In einer direkten Annäherung ein spezifisches Sprachpaar wird betrachtet, und Hauptgewicht wird auf ein zweisprachiges Verzeichnis als Wissen Quelle gesetzt. SYSTRAN's-Erfolg ist groß das Resultat der großen Bemühung, die innen über viele Jahre gesetzt worden ist, um linguistisches Wissen in sein System zu enthalten.

[back to structure of presentation]

Babel Fish Translations

http://babelfish.altavista.com/

free real-time translations on the Internet powered by SYSTRAN

"Translations normally appear more accurate going from a foreign language into English" (SYSTRAN Sample Translations, 1)

SL

SYSTRAN ist zur Zeit das am weitesten verbreitete maschinelle Übersetzungssystem. Es ist extrem erfolgreich, obwohl es ein frühes, einfaches Design hat. Das bedeutet, dass SYSTRAN ein typisches Beispiel ist für eine direkte Herangehensweise der ersten Generation, im Gegensatz zu einer interlingualen Herangehensweise der zweiten Generation und einer Transfer-Herangehensweise der dritten Generation. Davon haben wir letzte Woche gehört. Bei einer direkten Herangehensweise wird ein spezifisches Sprachpaar betrachtet, und der Schwerpunkt wird auf ein zweisprachiges Wörterbuch als Wissensgrundlage gelegt. SYSTRANs Erfolg ist zum großen Teil das Ergebnis der grossen Anstrengung, die über viele Jahre hineingesteckt wurde, um sprachwissenschaftliches Wissen in das System aufzunehmen.

TL

SYSTRAN is at present the furthest common machine translation system. It is extremely successful, although it has an early, simple Design. That means it that SYSTRAN is a typical example for a direct Herangehensweise of the first generation, contrary to a interlingualen Herangehensweise of the second generation and a transfer near going way of the third generation. Of it we heard last week. With a direct Herangehensweise a specific language pair is regarded, and the emphasis is put on a bilingual dictionary as knowledge basis. Systran success is to a large extent the result of the large effort, which was put in over many years, in order to take up linguistic knowledge to the system.

[back to structure of presentation]

Babel Fish Translations

http://babelfish.altavista.com/

free real-time translations on the Internet powered by SYSTRAN









ambiguous sentence

He saw her shaking hands. Er sah sie, Hände zu rütteln.

? He saw her trembling hands. Er sah sietrembling Hände.

? He saw her holding hands. Er sah ihreanhaltenden Hände.

ambiguous sentence

The man I want to leave. Der Mann, den ich gehen möchte.

? I want to leave the man. Ich möchte den Mann lassen.

? I want the man to leave. Ich wünsche den Mann gehen.

ambiguous sentence: does 'it' refer to 'vase' or 'window'?

She threw the vase at the window and it broke.

Sie warf den Vase am Window und es brach.


problem word 'put'

Did you put the book on the table?

Setzten Sie das Buch auf die Tabelle?

Did you put the book back on the shelf?

Setzten Sie das Buch zurück auf das Regal?

idiom

The student passed her test with flying colors.

Der Kursteilnehmer führte ihren Test mit Fliegen colors.


Kim gave the book to Robin. Kim gab dem Robin das Buch.

passive

? The book was given to Robin by Kim.

Dem Robin das Buch wurde von Kim.

?To sum up, Robin was given the book by Kim.

To aufsummieren, Robin wurde gegeben das Buch durch Kim.

fixed phrases

On the one hand, on the other hand. Einerseits, andererseits.

By the way.Übrigens. As far as I know. Insoweit ich weiß.

As far as I am concerned.Insoweit ich betroffen werde.

Translations normally appear more accurate going from a foreign language into English.


Übersetzungen erscheinen normalerweise genaueres Einsteigen von einer Fremdsprache ins Englische.

Es wird immer kälter. It becomes ever colder.

anaphoric pronoun

Ein Tisch hat 4 Beine, und manchmal hat er eine runde Tischplatte. Manchmal hat er auch eine rechteckige Tischplatte aus Glas.

A desk has 4 legs, and sometimes it has a round desk top. Sometimes it has also a rectangular desk top from glass.

[back to structure of presentation]

SYSTRAN - Past and Present:

A Brief History of SYSTRAN Translation Software

[back to structure of presentation]

How does SYSTRAN work?

summary of the article by Peter Wheeler 'Systran', 1987


 

[back to structure of presentation]

The Implementation of SYSTRAN at XEROX

summary of the article by John Elliston

'Computer Aided Translation: A Business Viewpoint', 1979









The Need

XEROX needs to communicate technical Service Documentations in 15 different languages to Service Representatives.

The Solution

XEROX implemented SYSTRAN as a tool to assist in the translation of technical Service Documentations from English into foreign languages.

The Requirements

The original Service Documentations (SL) must be written in a clear manner to avoid ambiguity.

? Input controls:

XEROX works with vocabulary lists (MCE list - multinational

customised English)

? Vocabulary used in SL text must be from the MCE list

Writing rules are imposed on SL, e.g. simple, short sentences
 
 

The Process generation of SL text (Service Documentation in English)

raw machine translation with SYSTRAN

post-editing

final printout of TL text
 

[back to structure of presentation]

The Implementation of SYSTRAN at the Nuclear Center of Karlsruhe (Kernforschungszentrum Karlsruhe)

summary of the article by F.W.A. Habermann

'Provision and Use of Raw Machine Translation, 1986



 
 
 
 
 
 
 
 

The Need

The Nuclear Center of Karlsruhe needs an MT system to translate research papers concerned with fast breeder technology and written in French.

The Solution

A French - English SYSTRAN MT system

? much faster than human translation

? limits human work to a minimum
 
 

The Restrictions

The application of the MT system is limited to scientific texts

Raw MT output is used without post editing

It is a French - English MT system only

The Process

SL text is scanned for spelling errors

SYSTRAN dictionary lookup

Listing of not found words in the "Not Found Word List"

Manual input correction with help of "Not Found Word List"

Run of SYSTRAN translation program

No post editing

Side by side print of French source text and English translated text

[back to structure of presentation]

The Role of the Posteditor:

summaries of the articles by

R. Green 'The MT Errors Which Cause Most Trouble To Posteditors', 1982

B. Lavorel 'Experience In English - French Post-Editing ', 1982

B. Kostrewski & P. Taylor 'Discussion on Session 3: The Posteditors'

Experience', 1982
What has the post-editor to take into consideration?

accuracy is the most important consideration

complete intelligibility may not be essential

in technical and commercial fields style is not first priority

Most post-editors are translators rather than revisors [...] , their work is assessed on the basis of accuracy and quality, perfection of style being one of the main criteria. [...] they are unwilling to hand in work written in any kind of standardized style [...] Potential post-editors should therefore be given an objective explanation of what the machine can and cannot (yet) do and why (Lavorel, 108) posteditors should always work with the TL text and the SL text

the quality of the final product that the post-editor seeks to achieve

may be quite different from what the user requires even if MT output may be nowhere near acceptable to human translators, it can nevertheless be acceptable to the user concentrated emphasis on detail in post-editing is a luxury which commercial concern cannot afford [back to structure of presentation]

BIBLIOGRAPHY


Elliston John S.G. "Computer Aided Translation: A Business Viewpoint". In: Translating and the

Computer. Ed. by Barbara M. Snell. Amsterdam: North-Holland Publishing Company, 1979:
149 - 158.
Green, Roy. "The MT Errors Which Cause Most Trouble To Posteditors". In: Practical Experience of Machine Translation. Ed. by Veronica Lawson. Amsterdam: North-Holland Publishing
Company,1982: 101 - 104.
Habermann, F.W.A. "Provision and Use of Raw Machine Translation". In: Terminologie et Traduction. 1986: 29 - 42. Kostrewski, Barbara and Peter Taylor. "Discussion on Session 3: The Posteditors' Experience". In: Practical Experience of Machine Translation. Ed. by Veronica Lawson. Amsterdam: North-
Holland Publishing Company,1982: 111 - 112.
Lavorel, Bernhard. "Experience In English - French Post-Editing". In: Practical Experience of Machine Translation. Ed. by Veronica Lawson. Amsterdam: North-Holland Publishing
Company,1982: 105 - 109.
Wheeler, Peter J. "Systran". In: Machine Translation Today: The State of the Art. Ed. by Margaret King. Edinburgh: Edinburgh Universitiy Press, 1987: 192 - 208. Whitelock, Peter and Kieran Kilby. Linguistic and Computational Techniques in Machine Translation System Design. London: UCL Press, 1995. Demo Versions

Free Translation.Online. Internet. 2.11.2001. Available HTTP: http://ets.freetranslation.com:5081/

Translate with Babel Fish. Online. Internet. 2.11.2001. Available HTTP: http://babelfish.altavista.com/

SYSTRAN: Homepage. Online Internet. 3.11.2001. Available HTTP: http://www.systransoft.com/

Web Sites

SYSTRAN: Machine Translation. Online. Internet. 2.11.2001. Available HTTP:

http://www.systransoft.com/MachineTranslation.html SYSTRAN: Past and Present. Online. Internet. 2.11.2001. Available HTTP: http://www.systransoft.com/PastPresent.html SYSTRAN Sample Translations. Online. Internet. 2.11.2001. Available HTTP: http://www.translation.net/syssamp.html SYSTRAN ® Translation Software. Online. Internet. 2.11.2001. Available HTTP: http://www.translation.net/systrans.html