![]() | Only 14 pages are availabe for public view |
Abstract The translation between different languages has an increasing importance due to the demand of information exchange, saving time of working people and technology transfer. This work presents a bi-directional Machine Translation (MT), an system for English-Arabic and Arabic-English translation of scientific domain. According to bi-directional MT, the system consists of two subsystems. Each subsystem runs in one direction (either English to Arabic or Arabic to English) using transfer-based approach and using two tightly related knowledge bases. Each subsystem consists of three main modules responsible for analysis, transfer, and generation. In the analysis components, the sentence is read and listed, the parser builds the grammar for this sentence by calling morphological analyzer and produces the syntactic tree relying on English/Arabic dictionary. The transfer component involves two steps: lexical transfer converts units in source text to equivalent in the target text with the same features, and structure transfer builds a target tree structure relying on set of rules. This process is carried out with the help of a bi-lingual dictionary. The generation component provides the target language (source verification text) translation, which involves the synthesis grammar rules of Arabic/English and an Arabic/English dictionary. The present work reports our attempt in developing the translation process by developing new rules that will enhance the translation process. This system is implemented in Prolog language. Experiments on real sentences were performed. The ability of this system enhanced the result process and reduced translation error of scientific texts from 14% to 10% by discovering new rules of translation based on bi-directional try-and-error translation. |