Analysis of pronunciation error and grammatical error of L2 learners for dialogue-based Japanese CALL system This paper describes a dialogue-based CALL (Computer Assisted Language Learning) system. One of the major problems in CALL systems is that learners are usually assigned a passive role. Learners have no practices in composing their own utterances. The other major problem is that lots of conventional CALL systems are pronunciation exercise systems. However, pronunciation exercise is only a part of exercise needed to increase a learner's communication skill. In this paper, we propose a dialogue-based CALL system that enables exercise learners' grammatical errors and lexical errors as well as pronunciation exercise. To construct proposed system, we have to solve lots of problems. When learners compose and pronounce various utterances, these utterances have many problems such as pronunciation errors, grammar errors (grammatical errors, lexical errors, out-of-grammar errors). To make such CALL system, we need to know what kind of mistake L2 learners make for pronunciation and grammar. Therefore, we investigate tendency of pronunciation from about 2800 utterances by 28 Korean adults, and we proposed pronunciation evaluation method [ 7 ]. This method computes an intelligibility score from a learner's utterance. Our work proved that the system's score correlates with an evaluation score given by native speakers. In order to solve grammar errors, especially out-of-grammar errors, we used a pre-exercise before doing conversation exercise with the system. Firstly, learners do pre-exercise such as vocabulary, grammar, and typical conversation examples. Then, they do real conversation exercises with the system. Because of the influence of pre-exercise, learner naturally compose sentences which are inside of the system grammar. However, we do not exactly know the influence and the effectiveness of the pre-exercise. We investigated the influence and the effectiveness of the pre-exercise. We also need to know what kind of grammatical and lexical errors L2 learners make in conversations. Therefore, we did interview experiments to find pre-exercise influences and grammar errors. According to the interview experiments, we know that the number of out-of-grammar sentences decreased if learners do pre-exercise. Therefore, by the influence of pre-exercise, learner naturally compose sentences which are inside of system grammar. In order to detect learners' grammatical and lexical errors, we examine several methods. According to recognition results, we check that proposed method (Sentence recognition rate by proposed method is about 80.6% and grammar error detection rate is 92.3% ) is effective. |