CORAA: a large corpus of spontaneous and prepared speech manually validated for speech recognition in Brazilian Portuguese.
Arnaldo Cândido JúniorEdresson CasanovaAnderson da Silva SoaresFrederico Santos de OliveiraLucas OliveiraRicardo Corso Fernandes JuniorDaniel Peixoto Pinto da SilvaFernando Gorgulho FayetBruno Baldissera CarlottoLucas Rafael Stefanel GrisSandra Maria AluísioPublished in: CoRR (2021)
Keyphrases
- speech recognition
- conversational speech
- automatic speech recognition
- brazilian portuguese
- speech retrieval
- speech signal
- speech synthesis
- broadcast news
- word error rate
- hidden markov models
- spontaneous speech
- machine translation
- language model
- speech processing
- speech recognizer
- noisy environments
- speech recognition technology
- pattern recognition
- recognition engine
- speaker identification
- speaker independent
- speech recognition systems
- handwriting recognition
- speech recognizers
- keyword spotting
- vocal tract
- speech recognition errors
- acoustic features
- speaker dependent
- word recognition
- parallel corpora
- maximum likelihood
- acoustic models
- cepstral coefficients
- speaker recognition
- background noise
- noisy speech
- linear prediction