This paper presents a prototype sentence-level system for Example Based Machine Translation (EBMT), which considers Arabic Language. Through the paper, it is shown that it is possible to have a machine translation system without an Arabic analysis module. Through the paper, we will show that the system is able to produce a good quality translations from few translation examples. The unit of translation is sentence level, which is proposed to be a suitable level when considering Arabic. The proposed system has two modes of operation: training phase and the translation phase. In training phase a parallel corpus is constructed automatically at the sentence level. The corpus is tagged using morphology-based part-of-speech assignment for both English and Arabic sides of the corpus. In the translation phase, the system depends on pattern-based syntactic matching of the input English sentence, and the pattern stored in the database. Once a matching occurs, the system gradually modifies the corresponding Arabic part to produce correct translation. The parallel corpus database is implemented in Access, and currently includes 1413 analyzed examples. The system is implemented in Visual Basic linked with the Access database.