Introduction Rapid changes in the global marketplace have given rise to new demands and have provided new opportunities for the translation industry. The need for multilinguality in the presentation and business logic layers of most modern systems, applications and services is a great challenge that the translation industry now faces. But even after many years of intense research and many commercial attempts of related products, translation systems of today still fail to completely meet the above needs, still need something more. Systems that integrate different technologies (eg Machine Translation & Translation Memory Engines), although performing better, they do not seem to solve the problem. The need for further development of existing translation engines through the acquisition of new intelligent tools and methods appears to be fundamental. Within this framework, much discussion about the architecture and constituency of modern machine translation systems is being made. Actually, two of the dominant, in the R&D world, translation technologies, statistical machine translation and example-based translation memories, share some drawbacks that could be uniformly overcome. Statistical, as well as traditional rule-based, systems are vulnerable in cases of multiple transfer equivalents. Such systems are usually deficient in selecting the appropriate translation according to statistical evidence within a general bilingual corpus. Translation Memory systems have also presented limited success with respect to the type and the size of the text units involved in the translation process. Since their performance relies heavily on the existence of ³good" matches they are characterised by considerable inflexibility and a rather ungraceful degradation curve when these matches are poor.
Given as input a source sentence to these translation systems, the approach proposed in this paper attempts to improve their performance in producing correct translations. The method expands the transfer selection capabilities of these systems even when a small bilingual parallel corpus is utilized. Regarding the adaptation to a Translation Memory Engine, in order to enhance the generative capacity (translation coverage) of the example database, a new type of resource was utilized within its framework: Sub sentence Translation Unit, which we consider as identical to the part of the input source sentence that has remained untranslated during the maching mechanism and in most cases contained a word, phrase or collocation. The real added value of any translation aid tool lies on its ability to encompass different levels of information and processing under a single framework towards providing optimal results. Thus, designing a language independent engine