If you are not familiar with Déjà Vu’s lexicon concept, get ready for something completely different and new.
You probably have a relatively good understanding of how the Termbase and the translation memory work. Both are project-unspecific, external databases that contain segments, phrases, and terms from a large variety of subjects, clients, and projects. While the translation memory is primarily used for segment matching, the Termbase provides the individual components of a segment when no match is found in the translation memory.
So how does the lexicon fit in?
The project’s lexicon is a list of all the source language words or phrases present in the project—in other words, an index of all terms and phrases. Once you have created this index, you can translate the terms that seem relevant to you, batch delete all other terms (for instance, terms like "the" or "to" in the image below), and Déjà Vu X3 will use what remains as the primary glossary for your project.
While creating this index, the lexicon also provides you with the number of occurrences of the individual terms or segments (see the Freq column in the image above), thus giving you a pretty good idea of what terms are central to your project. And even though the same terms may reside in your Termbase, chances are that there will be several translations for many terms in the database. The word "cat," for instance, could be a feline animal, a piece of heavy equipment, or "computer-assisted translation." Each of these meanings will have a different translation in different languages, making it difficult for Déjà Vu X3 to know which one to choose over the other. By providing the one correct translation in the project-specific internal lexicon, Déjà Vu X3 will always choose that over any other translation.