This interface provides on-line access to the LOGON Machine Translation demonstrator. Driving a rich cascade of computational grammars and specialized processing tools, the interface allows one to input one Norwegian sentence at a time, subsequently translate it to English in a stepwise fashion, and inspect intermediate results at varying levels of detail.
This proof-of-concept demonstrator implements a ‘deep’ linguistic approach to machine translation between Norwegian and English, where a given source-language input is analyzed into an abstract, logical-form semantics (using Minimal Recursion Semantics; MRS); which is subsequently transferred into a target-language semantic form; which in turn is used for full, grammar-based generation of the English system output. The approach capitalizes on translation quality and precision, rather than on broadest possible coverage or robustness. At its current development stage, the system only has access to a limited vocabulary (specifically in transfer and generation). Therefore, choose your input carefully and inspect intermediate results (the probably most interesting aspect of the demonstration) with great care for details. The LOGON target domain is on hiking instructions and general tourist information. Clicking the Sample button will activate a random selection from the LOGON target corpus of some 5,000 sentences.
The LOGON on-line pages make relatively heavy use of Cascading Style Sheets (CSS) and (some) JavaScript; while it should be possible to navigate without JavaScript, for the CSS to work properly, you will need a modern, HTML 4.0.1 compliant browser (e.g. FireFox or Mozilla 1.x, Netscape 7.x, IE 5.x, and later versions or derivatives). Please relay bug reports and suggestions for improvement back to the LOGON developers.