Auto-extraction of template library in template based machine translation(TBMT) system
Xiaodong Shi · Jisuanji gongcheng · 2004
TBMT need to include a large template library which must be built automatically. We presented a method to build the template library from the parallel corpora with setence aligned. First, we classified the sentences into several sentence sets by system cluster method so that the sentences in the same class had the same template. Then we calculated the similar degree of every two sentences according to the similar degree model. Last we extracted the template from every set. Thus the template library was built up.