Mining named entity transliteration equivalents from comparable corpora