발표문 텍스트의 어휘사용 특성: TED-LIUM과 BNC 코퍼스 비교

Abstract

This paper aims to present word frequency information of presentation transcripts for a better understanding of the linguistic nature of spoken data. The frequency information analyzed in this paper is based on the TEDLIUM corpus, which contains approximately 2.6 million words from 1,495 talks (transcripts) given by 1,242 different speakers. These talks are all task-oriented to some extent with various topics and communicative purposes, and they are of our potential interest as a preliminary to a measure of learning resources with the aim of developing listening and speaking skills. TTR (Type-Token Ratio) commonly used in evaluating lexical complexity or diversity of reading material is used to measure similar characteristics of presentation material as an exploratory measure. In addition, the frequency information extracted from TED-LIM is compared to the frequency list of BNC (British National Corpus) in order to understand similarities and/or differences between presentation transcripts and written/spoken corpus. Our better understanding of linguistic characteristics of spoken texts, though limited to presentation transcripts in this paper, can assist language learners with appropriate listening or speaking material to acquire effective communication skills.