Example-Based Machine Translation system EDGAR

Example-based machine translation (EBMT) in EDGAR consists of two steps. In a first step a bilingual reference text is aligned on a phrasal and word level and a biligual translation grammar is extracted from the aligned text. A new text can be translated based on the induced bilingual grammar and the aligned phrases and words.

These steps are described in a number of publications. EDGAR is used in a number of IAI projects for term recognition, term variant abduction and term control (pdf).
In a couple of experiments, EDGAR has been linked with the rule-based MT-system CAT2 and a translation memory.  These experiments are contained in an IAI electronic WP36.

Michael Carl has participated in the organization of a workshop on EBMT in September 2001 and edited a book "Recent advances in Example-based Machine Translation".

Publications