This project, which has just started, looks into how the assessment of the compositions written by Japanese language learners can be automated. We are concerned about the compositions submitted by students as part of their course assessment. That means, it is necessary to provide some feedback (not just a score for the composition), such as scores for different components (i.e. grammar, expressions) to assist the learning of students.

Research has been done towards the development of some systems for essay grading using state-of-the-art NLP techniques. In fact, there are some systems which are designed to do this kind of task in Japanese, but to our knowledge, none of them targets the compositions written by learners.

In this talk, we will present as to how we plan to progress this project (database, annotation, NLP techniques, etc), whenever applicable, mentioning some difficulties seemingly associated with the Japanese language.

Due to the complexity of its writing system and the significant differences between the spoken and written styles, much time needs to be spent on reading/writing in Japanese, compared to, i.e. many of the European languages. Naturally enough, composition is one of the most important parts of assessment in Japanese language learning. However, it is time-consuming and labour-intensive to mark compositions, which is adversative from a learners' point of view because they cannot receive prompt feedback. This is one of the incentives for this study.

We seek some collaboration with the expert in this area.

Authors: Shunichi Ishihara, Jun Imaki

Event: SF08: Speed Papers

← View all submissions for this event.