Method of extraction of transliterated-term pairs with iterative process

A method of extraction of transliterated-term pairs with iterative process is provided for employing an acoustics statistics model to create a confusion matrix; it comprises a method of automatic transliterated-term extraction, a method of automatic conversion of cross-linguistic syllable-and-phoneme and a calculated method of similarity degree between syllables for alleviating syllable variation. The acoustics statistics model of the present invention is chosen for speech recognition in great quantities to create a confusion matrix and acts as an initial characteristics value for transliterated-term extraction to alleviate the pronunciation variation problem; a cross-linguistic syllable-phoneme correspondences is trained by using the extracted term pairs, then iteratively extracting transliterated-term pairs according to the correspondences.