Hungarian Parliamentary Speech and Aligned Text Selection Database
Database of recordings and official transcripts of Hungarian parliamentary speeches. The recordings are segmented between speech pauses, which not necessarily correspond to sentence boundaries. The official transcripts are not completely accurate, since the parliamentary transcribers correct most of grammatical mistakes and speech disfluencies. Hence, an automatic speech recognizer was utilized to choose only those segments, where there is a high match between the automatic and manual transcriptions. Thus the database comprises only those segments that are considered to have a reliable transcription. The database can be applied in speech technology research, phonetic, phonological research and for developing and testing speech and speaker recognition systems.