Skip to content

Proper-noun / guest-name correction using episode metadata #6

Description

@frarredondo

Summary

ASR often mangles guest names and show-specific proper nouns that are already present in episode metadata.

Motivation / evidence

Examples from local transcripts:

  • Sabina Vosheshak ← Sabine Wojcieszak (title had the correct name)
  • Patrick Akyo, Uers Peter
  • Chain Street / JaneSheet ← Jane Street

Proposal

Build a per-episode glossary from title, show name, and RSS description/participants, then apply a conservative correction pass to cleaned text (and optionally report substitutions).

Out of scope

General spell-checking of the whole transcript, LLM rewriting of prose, changing timed segments unless explicitly designed.

Acceptance criteria

  • Names present in episode metadata are corrected when clearly misspelled in text
  • Conservative matching (avoid false positives on common words)
  • Can run via sync flags and/or podtx format
  • Docs / --help updated

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions