Machine translation and translation memory : Breaking the barriers [CD-ROM]

Editado en CD-ROM.

As explained in the introduction, current models (TM and MT) must unite. I do believe that machine translation must keep its central position as the rule-based computational foundation. But recent trends indicate that there is an increasing tendency to complement rule-based (or algorithmic) translation with increasingly vast corpuses: corpuses of expressions, of idiomatic constructions, specialised glossaries that are automatically loaded when a certain context is detected; and most of all, translation memories. My aim was to open a perspective on how to squeeze out as much as possible from translation memories, how to devise synthetic, or artificial, T-grammars, and show strategies for using databases of immense sizes by fractally compressing them.