Pixel International Conferences

In these days, many foreigners visit Japan from many countries because of sightseeing, education or getting jobs. In Japan, they have many natural disasters such as earthquake, tsunami, flood or eruption. When a disaster occurs, an authority is responsible to inform of various information for evacuation or life under a refugee camp to not only Japanese native speakers but also non-native speakers. However, multilingual announcement is not a realistic way of achieving this because of the limitation of human resource under such an emergency situation.

 Under this background, “Easy Japanese” (EJ) has been proposed [1]. EJ is a constrained language, which consists of limited vocabulary and grammar. Thanks to the linguistic limitation, a sentence in EJ is more easily understood than that in ordinary Japanese by Japanese non-native speakers. Because of its easiness to use, "Easy Japanese" has been widely used in public office web site, leaflet or broadcast.

 The problem is that composing sentences in EJ needs training, because a Japanese native speaker cannot understand what words or phrases are difficult to understand for non-native speakers. To make composition in EJ easier, we developed a system that helps a writer to compose sentences in EJ without knowledge of the limitation of EJ. The system is named YANSIS, which stands for “YAsashii Nihongo SIen System” (Easy Japanese writing support system).

 YANSIS consists of six components. The UI component provides user interface such as Japanese text input and buttons. The Japanese morphological analyzer component is used to split an input sentence into words. We need this component because a Japanese sentence do not have spaces between words. The Japanese level analyzer component determines difficulty level of each word based on the vocabularies of Japanese language proficiency test (JLPT). The recommendation component finds out difficult phrases in the input sentence and recommends how to rewrite them. The Japanese easiness estimator component estimates difficulty of the whole sentence based on machine learning technique [2]. The example search component searches examples of EJ sentences that are related to a word in the input sentence.

 The first version of YANSIS was implemented in Java, and thus it runs on any OS if Java runs on it, such as Windows or Linux or OS X. In addition, we ported YANSIS to Android and iOS. Because Android apps are based on Java, we could reuse many components of the original version in the Android version, except the UI component. However, for Java is not available on iOS, we needed to re-implement all components including Japanese morphological analyzer using Objective-C.

[1] “Easy Japanese”: http://human.cc.hirosaki-u.ac.jp/kokugo/EJ1a.htm

[2] M. Zhang, A. Ito and K. Sato, “Automatic Assessment of Easiness of Japanese for Writing Aid of “Easy Japanese” ”, Proc. Int. Conf. on Audio, Language and Image Processing, Shanghai, pp. 303-307, 2012.