Developing spoken language corpora is difficult because of the tremendous effort required for recording and transcription, and this has hindered the construction of large-scale spoken language corpora. Our project aims at developing a methodology for sharing existing conversation corpora that cover diverse styles and settings. As a first step in this endeavor, we examined the different transcription conventions for corpora that have been developed by various researchers, and then attempted automatic conversion between CSJ-style and CA-style transcriptions. The accuracy of our method was quite high, although there is still room for improvement.
〈共同研究プロジェクト紹介〉独創・発展型 : 多様な様式を網羅した会話コーパスの共有化 会話コーパスの共有化に向けて : 転記方式の自動変換
review040203.pdf
(553.4KB)
運用指針、オープンアクセス方針、等
