Name Romanizer

In multilingual applications, the problem of corss-lingual name matching is complicated by the fact that when a name is represented in a script different from its native script there may be several alternative representations for each phoneme, leading to large number of potential variants, this is true for Romanized forms of names in many languages such as Arabic. Romanizer does an orthographic transcription by default, a process that can distinguish between multicultural names and apply a sophisticated algorithm to provide the most likely variations. It is neither a simple name matching algorithm nor a name combination generator engine; it rather goes all the way back creating a new legitimate version of the input name token to decide which legitimate variants to return, thus avoiding the elaborate listing of trivial and never-exist variants mostly output by simple combination processes. It also tracks cultural variations through look-up database of refined set of etymologic database to follow exactly the way used by specific culture to Romanize the name.

For Semetic languages such as Arabic which is normally unvocalized, the unvocalized Arabic name does not give sufficient information for accurate romanization, Arabic personal names come without any diacritical marks so a fully accurate vocalized version of the name is necessary for romanization purposes, the system serves as a major pre-processor and provide three optional vocalization modes to chose from as adding vowels to Arabic personal names is critical for any Romanization or transcription process, this module accomplishes the task of disambiguating Arabic personal names; it can be viewed as a front-end processor for the transcription system, this module is name specific, unpredictable output will result if used to vocalize ordinary text

Arabic personal name has no more than one way to spell and this is true for most of the names exist in the language vocabulary; variants occur only when romanizing the Arabic name to other languages that have no one-to-one phonetic relationship with Arabic e.g. all romance languages. As a result of this each Arabic name has many romanized variants in other languages too.

A major problem challenging Arabic text processing systems is the absence of diacritics and vowels (Fat'ħa, Damma, Kasra, Shadda, Sukuun) in almost all the Arabic writings. The information contained in unvocalized orthography of the Arabic names is not sufficient to most applications supposed to accomplish automatic transliteration of Arabic proper nouns to English For instance the names "أحْمَد" /ʔħ'mad/ and "مُحَمَّد" /muħam'mad/ " have the following different transcription pairs respectively :

"Ahmet" and "Muhemet" in Turkey,
"Ahmed" and "Mohamed" in North Africa e.g. Morocco, Algeria, Tunisia, Mauritania, and Egypt
"Ahmad" and "Muhamad" in Gulf region e.g. KSA, Iraq, Oman, etc.

The pairs shown above do not actually tell the whole story since the list of variants of such very common names will be really long without enfocing some sort of rules to limit the output within acceptable phonemic forms. The system is designed to deal with such complexities and fully return all romanized versions of any personal name with such minimal nuances been distinguished and clearly manifested