作者: Dino Seppi , Matteo Gerosa , Björn W. Schuller , Stefan Steidl , Anton Batliner
DOI:
关键词:
摘要: In this paper we describe the eectiveness of some linguistic features for detecting problems in spoken child-computer interactions. To aim, use an Automatic Speech Recognizer generating word chain, and a tokenizer obtaining lexical stemming information. classification each turn is eventually achieved by exploiting frequencies tokens’ classes. The impact ASR tagger accuracy on automatic detection are discussed comparing fully with manually corrected approaches.