Abstract
Our joint research project of creating an original hypermedia corpus of spoken Japanese began in 1991 as a pilot study for error analyses in discourse. We have collected 'live' spoken data from actual conversation between experts of TJFL (Teaching Japanese as a Foreign Language) and several native and non-native speakers (or learners) of Japanese. The experiments were executed in the forms of free talking plus role plays, which were based upon an established testing format known as OPI (Oral Proficency Interview). The whole contents of OPI experimentation were first recorded by camcorders on several high precision video tapes, and then converted into digital video files and/or digital sound data files, using DTV hardware and software. So far, the accumulated data (compressed files) have amounted to more than 1GB, which include movie clips, sound files, and full texts transcribed from the original digital sound data. The process of text transcription and its link to video and audio data is still under way. Our research has also been motivated by a new concept of global network. Recently, a sharp rise of interest in communicating ideas and sharing resources via Intemet seems to be accelerating the move toward building up distributed databases, accessible to any user connected to the worldwide network, e.g. WWW, Gopher and WAIS. The sample movie data of our corpus project is now being transferred to WWW server (http://www.fit.ac.jp) along with HTML-based tagged texts and is expected to be available on the Internet.
