LogonReports

Overview

The LOGON system is scarcely blessed with technical documentation. However, for some of the invidicual components and specific aspects of the system, a collection of [http://www.emmtee.net/reports/ technical reports] provides some background reading. As of late 2008, however, many of these documents are in a relatively early stage of production, i.e. they may be incomplete or at times a little rough in form. The LOGON developers make this draft reports available in the hope that they can help interpret more of the technical details of the LOGON MT infrastructure. Critical feedback on these reports will be warmly welcomed by the authors.

English Resource Grammar

Dan Flickinger provides a summary of adapting the ERG for use in the Norwegian–English system configuration. LOGON [http://www.emmtee.net/reports/7.pdf Technical Report # 2007-7] is comparatively complete and stable.

Realization Ranking

In the context of Norwegian–English MT, Erik Velldal gives a condensed overview of his [http://www.velldal.net/erik/pubs/Velldal08.pdf PhD research] on ranking generation results, i.e. the sets of paraphrases produced by the final stage in the LOGON pipeline: target language realization from transfer outputs. LOGON [http://www.emmtee.net/reports/10.pdf Technical Report # 2007-10] is largely complete, although, being a perfectionist, Erik might eventually apply some further polishing (once he returns from his [http://sailingkajsa.net/ celebratory voyage] after his June 2008 PhD defense).

MRS-Based Transfer

Albeit very incomplete and sketchy in several respects, Stephan Oepen provides a semi-formal summary of the MRS transfer formalism developed in the LOGON project. LOGON [http://www.emmtee.net/reports/11.pdf Technical Report # 2007-11] should be read with a grain of salt, but will hopefully be extended in the foreseeable future to fill in the remaining gaps.

Open Source Machine Translation

Originally as a visiting researcher to the Norwegian LOGON consortium, Francis Bond (and colleagues at NAIST, NICT, and NTT) was the first to apply the LOGON infrastructure to another language pair, Japanese–English MT. LOGON [http://www.emmtee.net/reports/21.pdf Technical Report # 2007-21] reviews some of the design decisions and challenges in adapting the original LOGON technology to other language pairs.