2018-12-10-re-aligning-timed-text.md

Re-aligning Timed Text

• Status: being evaluated
• Deciders: Pietro, James
• Date: 2018-12-10

Context and Problem Statement

When the user edits the timed text the transcriptions might lose the timecodes information over time after a number of edits. Eg if deleting a paragraph or line and re-writing it from scratch these words might not have time information associated with it.

Can we find a straightforward way to either preserve or restore the time information associated with the words of the corrected transcription?

Decision Drivers

• easy to reason around
• not too computational intensive
• the re-alignment operation can be performed on the client side
• Approach with a flexible granularity, eg re-align the whole text or just a paragraph or sentence

Considered Options

1. re-sync using audio waveform (TTS)
2. Transpose time-codes from STT transcript
3. Interpolate time-codes (words within a sentence)
4. Use Levenshtein distance (sentence level)

• …

Decision Outcome

still being evaluated

re-sync using audio waveform (TTS)

[example | description | pointer to more information | …]

• Good, because [argument a]
• Good, because [argument b]
• Bad, because [argument c]
• …

[option 2]

[example | description | pointer to more information | …]

• Good, because [argument a]
• Good, because [argument b]
• Bad, because [argument c]
• …

[option 3]

[example | description | pointer to more information | …]

• Good, because [argument a]
• Good, because [argument b]
• Bad, because [argument c]
• …

Links

• [Link type] [Link to ADR]
• …