Development and Evaluation of a Multimedia Collocation Retrieval System for Language e-Learning as Contrasted to Conventional Learning Environment

Introduction Importance of collocation in efficient language learning is well recognized in the domain of second language education. In recent years investigations in brain science shows that retrieving from mental lexicon words acquired in conventional text-based learning is significantly slower on the scene of actual setting than those obtained in situationbased learning[1]. This paper adopts a broader definition of collocation in the web-based computer-aided multimedia language learning environment, and reports on an attempt of developing Multimedia Collocation Retrieval System (‘MultiCoReS,’ hereafter). for the learners of a second language, based on the mass supply of multimedia contents currently available on the web[2,3]. Our definition of multimedia collocation is based on the text data combined with speech and video information recorded synchronous to the text data; namely, collocation in the multimedia context. We argue that this broader definition of collocation entails the entire setting for the use of a particular expression that not only includes lexical and semantic usage but also supplies in multimedia format the paralinguistic context where it is actually uttered. This paper is also a novel attempt to effectively apply MPEG-7 technologies to retrieval of multimedia contents in language learning, enabling the user to facilitate the system and data portability. In order to clarify the effect of learning with MultiCoReS with a collocation database for learners of Japanese, we also report on a preliminary evaluation experiment using 10 Korean learners with the results that significant difference in the test performance were yet to be found between the two groups, i.e., one using MultiCoReS, and the other using conventional text-based method. The experiment with larger number of subjects is now underway.