Low-frequency words in bilingual corpora a step towards automatic extraction of bilingual word pairs

摘要

The high-frequency bilingual word pairs in bilingual corpora arc already listed in the dictionaries. It in the low-frequency pairs that we have to extract. Based on that idea, we examine the method for automatically extracting bilingual word pairs from corpora and show that the statistical method, which has been studied intensively so far, is not suitable for the task. If two words J1 and J2 which belong to the same language always co-occur in the same alignments, the statistical method cannot determine which word is the correct translation of word F which belong to the other language. we saw many of the low-frequency words are in the above situation.