DOMAIN WORD TRANSLATION BY SPACE-FREQUENCY ANALYSIS OF CONTEXT LENGTH HISTOGRAMS - Acoustics, Speech, and Signal Processing, 1996. ICASSP-96. Conference Proceedings., 1996 IEEE Inte

We report a new statistical feature relating a bilingual word pair in a non-parallel English-Chinese corpus. It is found that the lengths of context segments of a word is closely correlated to that of its translation, even when the corpus is non-parallel, i.e., monolingual texts which are not translations of each other. The context segment length histogram of a word has a characteristic pattern and corresponds to that of its translation. If a word appears most frequently in long segments, its translation is found to be most likely occurring in long segments. One way to match these histograms is to first extract their salient shape characteristics by space-frequency analysis and then match them against each other using dynamic time warping. The results of matching can be used in combination with other statistical features to bootstrap a word or term translation algorithm from non-parallel corpora. some previous approaches is their orientation toward European language pairs. They cannot be applied to language pairs such as Chinese and English. A new approach is which would be extendable to other language pairs is needed. This paper demonstrates a pattern matching method by using a statistical feature, the context length histogram, to correlate pairs of translated words. It will also be shown how space-frequency analysis is used for matching such word pair signals for translation. As input corpus, the bilingual transcription of the Hong Kong Legislative Council debates is used for experiments [6]. The data is from 1988-1992, with the first 73618 sentences from the English text, and the next 73618 sentences from the Chinese text. There are no overlapping sentences between the texts. The topics of these debates focus on the political and social issues of Hong Kong. 2. ALGORITHM OVERVIEW