A factor automaton approach for the forced alignment of long speech recordings

50 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech (2019, methodology): “The uploaded transcript, containing no timing information, is force-aligned (“auto-sync’ed”) using standard ASR algorithms to generate start and end times of each word [41].”
Lightly supervised recognition for automatic alignment of large coherent speech recordings (2010, methodology): “Other methods proposed in the literature use weighted finite state transducers (WFST, e.g. [3], [4]) or factor automata [5].”
Large scale deep neural network acoustic modeling with semi-supervised training data for YouTube video transcription (2013, methodology): “The transcript, containing no timing information, is force-aligned (“auto-sync’ed”) on servers using standard ASR algorithms to generate start and end times for reasonably spaced chunks of audio, over the course of the video [10].”
Automatic Building of Synthetic Voices from Audio Books (2010, methodology): “Between the islands of confidence, the Viterbi algorithm is employed to force-align the speech with the original text to obtain word/phone level boundaries [Moreno and Alberti, 2009], [Moreno et al.”