A New Confirmation Framework for Extracting Synonymous Chinese Transliteration Based on Pronunciation Model

Abstract

There is no transliteration standard across all Chinese language regions, including China, Hong Kong, and Taiwan, and variations in Chinese transliteration have thus arisen in the process of transliterating foreign languages (English, for instance) into the Chinese language. In this paper, we propose an integrated confirmation framework to confirm a pair, that is, a transliteration and another term, whether it is synonymous. This framework includes two major steps. First, we study methods from several pronunciation-based approaches to measure similarity between Chinese characters; these approaches are specified for the comparison of Chinese synonymous transliterations. Second, we construct a new confirmation framework to confirm whether a pair of a Chinese transliteration and another Chinese term is synonymous. The presented framework is applied to extract synonymous transliteration pairs from a real-world Web corpus; this is valuable to build a new database of synonymous transliterations or support search engines so that they can return much more complete documents as Web search results to increase the usages in practice. Experiments show that our integrated confirmation framework is effective and robust in confirming and extracting pairs of Chinese transliteration following the collection of synonymous transliterations from the Web corpus.