تمثيل المعجم الدلالي باستخدام الذاكرة الارتباطية ثنائية الاتجاه لدعم الترجمة الآلية

الملخص EN

In this paper a neural network based translation system is presented.

Feed forward architecture is adopted to model the association scheme of the proposed system.

By this architecture, the Bi-directional association memory is designed to fulfill the requirement required to represent the lexicon dictionary in retrieving meaning of data, in two possible languages.

Translation system depends on relevant expressions to represent the data base.

In this data base a set of files is organized.

Each file is assigned for specific task to store classes, meaning, roots, predecessors, successors and expression formats.

By this organization terms are enabled to be thoroughly attributed to it’s contextual, grammatical and syntactical characteristics.

Vectors are used to represent data in order to support the encoding process in powerful data compression scheme.

For the sake of explaination, different examples are used.

Through out the study of these examples, the efficient attributes of neural networks are proved to be applied successfully into the proposed application of the presented translation system.