Why a transcript library is not yet a knowledge system

Text preserves only part of the record

A transcript makes speech searchable, but a sentence in isolation can lose the speaker, timing, surrounding conversation or uncertainty in recognition. A summary adds another transformation. Keep each derivative connected to the material it was created from.

Consider a constructed example: a speaker says ‘we might ship on Friday’, while a summary says ‘shipping Friday’. The summary has removed uncertainty. Searching for the date alone cannot reveal that change; a link back to the passage and recording can help a reviewer inspect it.

A useful passage needs context

A practical record might include a recording identifier, transcript version, passage offsets or timestamps, speaker label where reliable, extraction method and review status. Stable identifiers let a correction propagate without treating every derivative as unrelated text.

W3C PROV-O supplies relationships for derivation and attribution. The application design proposed here uses those concepts without requiring every small system to adopt RDF.

• Retain the permitted source recording and its identity.
• Version transcript corrections instead of overwriting their history.
• Link generated summaries and claims to supporting passages.
• Mark uncertain speaker labels and recognition errors explicitly.
• Review consequential claims against the original context.

Search is access, not validation

Retrieval can locate a plausible passage; it cannot establish that a summary preserved its meaning. Keep the distinction between a search result, an extracted claim and an accepted record. A reviewer should be able to reject a claim while retaining the evidence that prompted it.

Wisprs is AEY's publicly disclosed audio/video transcription product. Its public site describes transcripts, summaries and exports. This article proposes a broader knowledge-record design; it does not claim that Wisprs implements every control described here.

Limits and preparation

This is a source-informed architecture insight, not an evaluation of transcription accuracy or a product benchmark. The example is synthetic. AI assisted preparation; the company owner approved editorial/disclosure publication on 3 October 2026.

Sources

• W3C PROV Ontology
• Wisprs public product description