Corpora compilation for prosody-informed speech processing

Abstract Research on speech technologies necessitates spoken data, which is usually obtained through read recorded speech, and specifically adapted to the research needs. When aim deal with prosody involved in available data must reflect natural conversational costly difficult get. This paper presents a machine learning-oriented toolkit for collecting, handling, visualization of using prosodic heuristic. We present two corpora resulting from these methodologies: PANTED corpus, containing 250 h English TED Talks, Heroes corpus 8 parallel Spanish movie speech. demonstrate their use deep learning-based applications: punctuation restoration translation. The presented are freely community.