To create a parallel and precisely annotated multispeaker speech database was the goal. Its precise marking and labeling background provides a solid basis for speech research, scientific investigations, and speech technology developments. Parallel character means that the 5 women and 5 men read the same Hungarian set of sentences (2000) under the same recording conditions. The precision indicator means that the annotation, tagging and other data parallel with the speech wave are highly precisely given, created by a combination of machine and manual processing. The manual check means that every sentence (every speech sound) of the database was subjected to a visual and auditory check, and if any corrections were necessary, they were implemented. The end result is PPBA. The entire database contains 6x2000=12000 sentences. This is the only Hungarian speech database with complex acoustic and linguistic content. More details can be found in the scientific paper in Hungarian.
Structure of the Hungarian Parallel Speech Database.
One of the sentences of the database read by the 10 speakers