Gradable Quality Translations through Mutualization of Human Translation and Revision, and UNL-Based MT and Coedition

Translation of specialized information for end users into many lan- guages is necessary, whether it concerns agriculture, health, etc. The quality of translations must be gradable, from poor for non-essential parts to very good for crucial parts, and translated segments should be accompanied with a meas- ured and certified "quality level". We sketch an organization where this can be obtained through a combination of "mutualized" human work and automatic NLP techniques, using the UNL language of "anglosemantic" graphs as a "pivot". Building the necessary multilingual lexical data base can be done in a mutualized way, and all these functions should be integrated in a "Montaigne" environment allowing users to access information through a browser and to switch to translating or postediting and back.

[1]  Christian Boitet,et al.  Coedition to Share Text Revision across Languages and Improve MT a Posteriori , 2002, COLING 2002.

[2]  Vincent Berment,et al.  Méthodes pour informatiser les langues et les groupes de langues « peu dotées ». (Methods to computerize "little equipped" languages and groups of languages) , 2004 .

[3]  Wang-Ju Tsai.  La coédition langue UNL pour partager la révision entre langues d'un document multilingue. (Coedition Language UNL to share the postedition among languages in a multilingual document) , 2004 .

[4]  Muriel Vasconcellos,et al.  SPANAM and ENGSPAN: Machine Translation at the Pan American Health Organization , 1985, Comput. Linguistics.

[5]  Etienne Blanc.  From Graph to Tree: Processing UNL Graphs using an Existing MT System , 2001 .

[6]  Bernard Vauquois,et al.  Static Grammars: A Formalism for the Description of Linguistic Models , 1985, TMI.

[7]  Christian Boitet.  A roadmap for MT : four « keys » to handle more languages, for all kinds of tasks, while making it possible to improve quality (on demand) , 2002 .

[8]  José Coch,et al.  Interactive Multilingual Generation , 2001, CICLing.

[9]  Christian Boitet,et al.  Representation trees and string-tree correspondences , 1988, COLING.

[10]  Christian Boitet,et al.  A research perspective on how to democratize machine translation and translation aids aiming at high quality final output , 1999, MTSUMMIT.

[11]  Christian Boitet,et al.  Four technical and organizational keys to handle more languages and improve quality (on demand) in MT , 2001, MTSUMMIT.

[12]  Christian Boitet,et al.  On UNL as the future “html of the linguistic content” & the reuse of existing NLP components in UNL-related applications with the example of a UNL-French deconverter , 2000, COLING.

[13]  Igor Boguslavsky,et al.  Creating a Universal Networking Language Module within an Advanced NLP System , 2000, COLING.

[14]  Yusoff Zaharin,et al.  Strategies and Heuristics in the Analysis of a Natural Language in Machine Translation , 1986, COLING.

[15]  Christian Boitet,et al.  A Rationale for Using UNL as an Interlingua and More in Various Domains , 2005 .