Arabic Transliteration

Recently, an increased interest in researches and experiments conducted for Arabic language has been seen. Some researches were related to segmentation and stemming of Arabic words (Chen & Gey, 2002; Xu et al., 2001; Lee et al., 2003; Rogati et al., 2003). Other researches dealt with cross-language information retrieval for query translation based on several tools such as machine translation systems and bilingual dictionaries (Chen & Gey, 2001; Fredric et al., 2001; Franz & McCarley, 2002; Darwish & Oard, 2002). It is valuable to create multilingual dictionaries and to be able to successfully distribute them (Lafourcade, 1997). The approaches that were presented by Kay & Roscheisen (1993) and Stanley, did not produce as comprehensive

One way to deal with proper nouns when the two languages under consideration have different alphabets is to transliterate these words. That is, to render them in the orthography of the other language. Transliteration is the process of representing of words in one language using the alphabet of the other language. There are some challenges facing the English-Arabic transliteration process. An Arabic name may be written in English with some different spellings. For example, the name Sulayman سليمان)) is written only one way in Arabic. This name is written in many different ways in English, such as Salayman, Seleiman, Solomon, Suleiman, and Sylayman. Moreover, further variations perhaps occur in the transliteration of proper names, since the owners of the names may take certain liberties with both the spelling and pronunciation of their names, for either devious or aesthetic reasons. Aside from the problems mentioned above that are inherent in transliteration, the transliteration of Arabic has an additional challenge: Words are written without “short” vowels. As a result, the transliteration process should include a vowelization step, in which the appropriate vowels are inserted into the un-vowelized proper noun (Arbabi et al., 1994). A traditional proper noun is another challenge. The term traditional proper noun is used to denote foreign

Proper nouns and corresponding transliterations may be found in parallel corpora. However, many approaches dealt with this problem based on the frequencies of words appearing in corpora. This approach may not be effectively applied to low-frequency words, such as transliterated words (Fattah et al., 2006a). Fung, used some approaches to create translation pairs from comparable and parallel corpora (Fung, 1995; Fung & Yee, 1998; Fung, 1998). For example, in (Fung, 1995), she presented a pattern matching approach for compiling a bilingual lexicon of proper nouns and nouns from unaligned parallel texts of Asian/Indo-European language pairs. The approach was simple but the recall was very small. On the other hand, (Fattah et al., 2006a) presented two algorithms and their combination to automatically create an Arabic/English bilingual dictionary from texts that exist in the Internet archive after using an Arabic light stemmer as a preprocessing step. Both Fattah and Fung approaches do not require pronunciation dictionary to convert a source word into a target one and they give reasonable