| Computers have long been regarded as a potential tool for accelerating the learning of a foreign language. Unfortunately, many attempts to incorporate computers into foreign language instruction have met with failure. One obstacle has been the relative 'primitiveness' of the existing Computer-Assisted Language Instruction (CALI) systems. Such systems typically provide mechanical drills involving fill-in-the-blank or multiple-choice questions, and utilize a "wrong-try-again" approach to interaction that offers little corrective feedback.;In this study, I developed a new CALI system, called "Nihongo-CALI". This system employs one type of artificial intelligence approach, Natural Language Processing, to facilitate more sophisticated error analysis and feedback. The Nihongo-CALI system uses Natural Language Processing to analyze the learners responses in light of the morphological and syntactic rules of the Japanese language, and provides feedback explaining the errors in terms of grammatical functions and thematic roles of the sentence.;The present study investigates the effect of the type of CALI exercises (traditional CALI exercises involving conventional feedback or intelligent CALI exercises providing sophisticated feedback) on students' language achievement and attitudes toward CALI exercises. The Nihongo-CALI system was implemented for both traditional CALI and intelligent CALI exercises, to process and to analyze the learner's input, but provided only limited feedback in simulating the traditional CALI approach. An experiment was performed to evaluate the comparative effectiveness of these two types of Nihongo-CALI exercises. Thirty-four subjects in the second-year Japanese language course at the University of Pittsburgh participated in the experiment.;The study found a statistically significant difference between the intelligent CALI and traditional CALI groups in the learners' achievement on the target structures, favoring the intelligent CALI group. The study suggests that the traditional feedback is as good as the intelligent feedback in helping learners to correct word-level errors (e.g., vocabulary and conjugation errors), while the intelligent feedback is more helpful for understanding and correcting sentence-level errors (e.g., particle errors). |
