v0.3.0 Release Notes¶
Release Date: November 17, 2026
Highlights¶
- Quotations checked against recordings: three content rules for what a comparison of several speech-to-text engines can't settle about a quotation from a podcast, interview or talk.
- Narration findings:
NarrationIssuefor a recording, such as a course voice-over, that doesn't say its script.
Quotations from recordings¶
A quotation attributed to a recording is checked by cutting a clip around it and comparing what several engines hear. When they can't settle it, the checker raises one of these findings on the claim:
| Rule | Problem |
|---|---|
content.quote.audioDisputed |
The engines disagree on a substantive word: a number, a negation, a name or a term |
content.quote.audioMismatch |
The engines agree with each other but not with the quotation |
content.quote.audioUnchecked |
Too few engines produced a transcript, or the clip could not be cut |
The existing content.quote.notFound applies when the quotation can't be located in the recording. Put what each
engine heard, and the clip's times, in the finding's evidence.
Narration¶
NarrationIssue is a recording, or one segment of it, that doesn't say its script,
raised as content.narration.mismatch. It projects to a finding in the content domain with type narration:
- the segment is the location:
Location.Componentis the segment ID, such as a slide's ID,Location.Filethe script's document andLocation.Snippetthe expected wording; - what the engines heard (
Heard) becomes atranscriptevidence entry; segment:<id>andrecording:<file>tags support filtering, andcontent.Segment(f)reads the segment back.
Changed¶
content.ClaimID(f)returns""for content findings of a type other thanclaim, so a narration segment is never read as a claim. Content findings with no type, as earlier versions wrote them, are still read as claim findings.
Compatibility¶
New constants and types only, plus the ClaimID change above. Claim findings written by v0.2.0 project and read back
unchanged.