Skip to content

Industry field guide

Television caption track revision data for AI

A broadcast program can have several versions with different caption tracks.

2022 NAICS
516120
Updated

What this guide tests

Which variant is CA checked against, and what apparent B alignment gap can be established from the supplied timing map?

See the specimen and fields · Work through the fictional records

Task and source context

A broadcast program can have several versions with different caption tracks. FCC accessibility materials and EBU media metadata provide context. A station can test caption identity and timing using an original training clip and two fictional transmission records. FCC accessibility reference · EBU media resource metadata

Create two practice variants with the same title and different start markers. Attach a checked caption track to one and an older track to the other. Ask which track the record supports, whether an apparent timing gap can be verified, and what must go back to an operator. Use a short independently written script.

Do not share network feeds, audience records, private ad schedules, full program masters, or rights-restricted captions. Clear both caption and audiovisual rights before actual reuse.

Describe a broadcast caption benchmark

Industry research

The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.

516120 · Television Broadcasting Stations

This industry comprises establishments primarily engaged in broadcasting images together with sound. These establishments operate television broadcasting studios and facilities for the programming and transmission of programs to the public. Programming may originate in their own studio, from an affiliated network, or from external sources.

U.S. Census Bureau, 2022 NAICS Manual, p. 412 ↗

Build a pilot sample

Set up this pilot

Make two original video cuts and two caption tracks with different start markers. Grade track assignment and line timing after converting the stated timebase.

Prepare the test packets

  1. Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
  2. Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
  3. Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.

Fields the reviewer needs

Program variant and track ID
Preserves which source the review covered.
Relative start marker and timebase
Makes the alignment calculation explicit.
Assignment versus completed review
Separates a configured track from a checked one.

Use Make a segment-level rights inventory for media and publishing to turn the specimen into a brief with a defined task, source and review owner.

Fictional exercise

A correct caption track can target the wrong program start marker

An original script is represented by two fictional program variants and simple seconds-based caption timings. No network feed, full master or restricted caption file is included.

Records supplied to the evaluator
Record IDSourceWhat it records
TC1Variant A and checked track CATitle Harbor-A; first spoken line at relative second 4. CA cues that line at second 4.
TC2Variant B recordSame public title, added 3-second opening; first line at relative second 7. No other timing changes documented.
TC3Transmission practice logB assigned old track CA; no B-specific caption review or corrected export attached.

Question for the model. Which variant is CA checked against, and what apparent B alignment gap can be established from the supplied timing map?

Show the reference answer

TC1 checks CA against variant A, with the first line at second 4. TC2 explicitly moves that line to second 7 in B. TC3's assignment of unchanged CA to B therefore creates a documented three-second early cue for that line under the supplied map. This establishes an alignment issue in the exercise, not a verified corrected track or a broader accessibility-compliance conclusion.

Download the fictional records and answer key (JSON)

How to check the result

Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.

Errors to flag in this case

  • Matching caption tracks on public title while dropping variant identity.
  • Treating a calculated offset as evidence that an operator has produced and checked a corrected export.

Change the inputs and run it again

Add an internal cut to B after the first line. Ask the model to keep the first-line offset while requesting a fuller map for later cues.

Score the decision and the evidence separately

Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.

Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.

Check permission and scope

Use an independent script and dummy logs. Exclude feeds, audience records, ad schedules, commercial masters and rights-restricted captions.

Keep guest, music, syndication and platform permissions separate. Exclude third-party audio or identifiable footage with unresolved authority.

Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.

Sources and further reading

Resources for this workflow

How the examples work

Company and program reviews

Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.

All company reviews

Related industry examples

← Information and publishing