Paper short abstract:

"AI Teacher" is an innovative and interactive language learning tool, and it can provide real-time corrections for students' grammatical errors. This paper discusses AI Teacher's characteristic features. It also reveals an intriguing fact about how language teachers correct students' errors.

Paper long abstract:

This paper purses the following two goals: (i) exploring how we can utilize language technology for language teaching and (ii) promoting the importance of linguistic training for language teachers.

First, we tackle the question of how we can utilize technology (e.g., natural language processing [NLP]) for language teaching. To this end, we present a system called "AI Teacher", which is currently under development. The AI Teacher is designed to detect students' grammatical errors and correct them automatically, using NLP. In order to acquire Japanese grammar rules, we crowdsourced a data set that consists of 12,000 error<->corrected pair sentences, such as (1) below from Japanese language teachers.

(1) [error]この本は、高いだと思います。 [corrected]この本は、高いと思います。

We successfully analyzed these data with the help of NLP, and extracted a large number of grammar rules for the AI Teacher's system. Here, we explain what types of grammar rules we have acquired from the data.

Our second goal is to address the question of whether we correct our students' errors consistently. This question arose when examining the crowdsourced data, mentioned above. In examining the pair sentences such as (1), we found that there was a wide range of variance in teachers' corrections of students' grammatical errors. We provide qualitative as well as quantitative analyses of our data. We particularly investigate the question of what linguistic factors would trigger such variance among language teachers, and why.

This paper not only showcases the AI Teacher (an innovative and interactive language learning tool) but it also highlights potential collaboration area(s) between technology and language teachers. Through the discussion of teacher variance, the paper also sheds new light on the field of teacher training.