Hebrew to English Machine Translation

We describe a hybrid transfer-based system for Hebrew to English Machine Translation. This task is particularly challenging due to two reasons: the high lexical and morphological ambiguity of Hebrew and the dearth of available resources for the language. We have used existing, publicly available resources which we adapted for this task. The system is based on a transfer engine which produces a lattice of possible translations and a statistical decoder which chooses the most likely translation according to an English language model. A small manually crafted set of transfer rules suffices to produce legible translations. The results are evaluated using state of the art measures and are shown to be encouraging. Machine translation of Hebrew is complicated due to two main reasons: the high lexical and morpho- logical ambiguity of Hebrew and its orthography; and the paucity of available resources for the lan- guage. In this paper we describe the first machine translation system for Hebrew. We used existing, publicly available resources which we adapted for this task. The system is based on a transfer en- gine (ref???) which produces a lattice of possible translations and a statistical decoder (ref???) which chooses the most likely translation according to an English language model. We manually designed a small set of transfer rules which reflect the most common local syntactic differences between He- brew and English. This small set turns out to be suf- ficient for producing legible translations. The results are evaluated using state of the art measures and are shown to be encouraging. In the next section we list some facts about the language, indicating sources for ambiguity. Sec- tion 3 describes the structure of the MT system with an emphasis on the resources required for its con- struction. Section 4 provides some translation ex- amples and a thorough evaluation of the system. We conclude with directions for future research.