Skip to Main Content
In natural language processing, quotes are an important grammatical category which needs consideration. In this paper, we propose a Japanese language model that includes quotes as a category. The quotes are recognized by using surface information and dependencies between the words. Then, they are divided into direct and indirect speech. Finally, we extract the quotes and create a relation between them and the original text. After the text has been analyzed, we obtain a tree structure with all the elements hierarchically categorized. We have experimentally tested the accuracy of the parsing process by creating a prototype system. The results show a 67.29% overall correct quote detection.