Japanese sentence compression using Simple English Wikipedia

We describe a cross-lingual approach for sentence compression of articles of Japanese Wikipedia using the correspondence of articles of Simple English Wikipedia. Taking advantages of the nature of the corpus, we can find essential parts from encyclopedic description without highly depending on the statistical information which are noisy. We manually explored the correspondences between the articles of Japanese Wikipedia and those of Simple English Wikipedia and then proposed a cross-lingual alignment method using simple matching algorithm. We provide an analysis of the abovementioned correspondence and the preliminary result of sentence compression using Simple English Wikipedia.