<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0712">
  <Title>An Integrated Approach for Arabic-English Named Entity Translation</Title>
  <Section position="3" start_page="87" end_page="87" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> The Named Entity translation problem was previously addressed using two different approaches: Named Entity phrase translation (which includes word-based translation) and Named Entity transliteration. Recently, many NE phrase translation approaches have been proposed. Huang et al.</Paragraph>
    <Paragraph position="1"> (Huang et al., 2003) proposed an approach to extract NE trans-lingual equivalences based on the minimization of a linearly combined multi-feature cost. However this approach used a bilingual dictionary to extract NE pairs and deployed it iteratively to extract more NEs. Moore (Moore, 2003), proposed an approach deploying a sequence of cost models. However this approach relies on orthographic clues, such as strings repeated in the source and target languages and capitalization, which are only suitable for language pairs with similar scripts and/or orthographic conventions.</Paragraph>
    <Paragraph position="2"> Most prior work in Arabic-related transliteration has been developed for the purpose of machine translation and for Arabic-English transliteration in particular. Arbabi (Arbabi et al., 1998) developed a hybrid neural network and knowledge-based system to generate multiple English spellings for Arabic person names. Stalls and Knight (Stalls and Knight, 1998) introduced an approach for Arabic-English back transliteration for names of English origin; this approach could only back transliterate to English the names that have an available pronunciation. Al-Onaizan and Knight (Al-Onaizan and Knight, 2002) proposed a spelling-based model which directly maps English letter sequences into Arabic letter sequences. Their model was trained on a small English Arabic names list without the need for English pronunciations. Although this method does not require the availability of English pronunciation, it has a serious limitation because it does not provide a mechanism for inserting the omitted short vowels in Arabic names. Therefore it does not perform well with names of Arabic origin in which short vowels tend to be omitted.</Paragraph>
  </Section>
class="xml-element"></Paper>