Synchronizing speech and text

Question

I have a text and a narration of the exact same text. What is the best way to synchronize them together? By synchronizing I mean, finding out for example the location of each word in the audio. For example if the sentence is "I took a cab" I want for markers on the accompanying audio which indicate start of each word.

Obviously any speech-to-text algorithm can be applied here, but given the fact that we've already know the text, I wonder if there's an algorithm for this simplified problem which works either much faster, or gives much better performance. Or if there is a general algorithm that can be tuned for this specific problem.

Answer 1

Papers on algorithms:

• A RECURSIVE ALGORITHM FOR THE FORCED ALIGNMENT OF VERY LONG AUDIO SEGMENTS Pedro J. Moreno, Chris Joerg, Jean-Manuel Van Thong and Oren Glickman

• Automatic Alignment and Error Correction of Human Generated Transcripts for Long Speech Recordings Timothy J. Hazen

• SailAlign: Robust long speech-text alignment Authors Katsamanis, Athanasios Black, Matthew Georgiou, Panayiotis Goldstein, Louis Narayanan, Shrikanth S.

Implementations:

• Sphinx4 long audio aligner

• Sail align

• Gentle forced aligner