LogonReports

Overview

The LOGON system (see the LogonTop page for background) is scarcely blessed with technical documentation. However, for some of the individual components and specific aspects of the system, a collection of 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 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 PhD research on ranking generation results, i.e. target language realization from transfer outputs. LOGON Technical Report # 2007-10 was not quite completed before Erik embarced on a celebratory voyage after his June 2008 PhD defense). When in doubt, the full PhD thesis may be a more reliable reference.

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 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 Technical Report # 2007-21 reviews some of the design decisions and challenges in adapting the original LOGON technology to other language pairs.