Improvement of Translation Accuracy for the Outlines of Japanese Statutes by Splitting Parenthesized Expressions

To globally share Japanese legal information, we translate the Outlines of Japanese statutes. These outlines are the official summaries of Japanese statutes and are useful to quickly understand their contents. In a previous statistical machine translation system for the outlines, we found that the training corpus consisted of both statutes and their outlines, including many long sentences that reduced the translation quality. To solve this problem, we shortened the length of sentences and focused on parenthesized expressions. In this paper, we propose a translation method that splits off parenthesized expressions from the sentences. Experimental result shows the effectiveness of our method.