Industry field guide
Court reporting transcript correction and citation data for AI
A search system that retrieves a prior transcript version may quote text that a reporter later corrected.
- 2022 NAICS
- 561492
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
Which source supports the current quoted meeting date, and does the index establish what E1's calendar says?
See the specimen and fields · Work through the fictional records
Task and source context
A search system that retrieves a prior transcript version may quote text that a reporter later corrected. U.S. Courts publishes federal court reporting guidance. A stenotype service can demonstrate citation and correction behavior using an original fictional hearing and invented page references. U.S. Courts court reporting guidance
Draft a short mock transcript, a separately identified corrected page and an index. Change a single date in the correction and keep the earlier page accessible to the evaluator. Ask which version supports the date and whether the supplied exhibit index establishes the exhibit’s contents. The answer key should point to the page and line in the correct revision.
Keep actual proceedings, sealed materials, client files and real case references out. If you add performance audio, secure written rights from each actor and review text and audio reuse separately.
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.
561492 · Court Reporting and Stenotype Services
This U.S. industry comprises establishments primarily engaged in providing verbatim reporting and stenotype recording of live legal proceedings and transcribing subsequent recorded materials.
Build a pilot sample
Set up this pilot
Write two original transcript versions and a correction limited to one line. Include a referenced but unavailable exhibit. Score the current quote without fabricating exhibit content.
Prepare the test packets
- 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.
- 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.
- 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
- Transcript version, page and line
- Makes the quote traceable.
- Correction target and replacement text
- Defines supersession.
- Exhibit reference and content availability
- Limits corroboration claims.
Use Build an AI evaluation sample for service handoffs to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A corrected transcript line controls the date while an exhibit index does not reveal exhibit contents
An independently authored fictional hearing fragment uses invented pages and no actual case or proceeding. Any later performed audio would require separate actor rights.
| Record ID | Source | What it records |
|---|---|---|
| CT1 | Original page 4, line 8 | Original mock sentence: 'The meeting was on day 12.' Transcript version T1. |
| CT2 | Corrected page 4, line 8 | Version T2 explicitly replaces T1 page 4, line 8 with 'The meeting was on day 14.' Other lines unchanged. |
| CT3 | Exhibit index | Lists exhibit E1 as 'calendar' with page reference only. Exhibit itself not supplied; no statement of its entries. |
Question for the model. Which source supports the current quoted meeting date, and does the index establish what E1's calendar says?
Show the reference answer
CT2 explicitly replaces the date at page four, line eight with day 14, so cite T2 and that location for the current transcript text. CT1 remains the superseded original. CT3 shows that an exhibit labeled calendar was indexed, but supplies none of its contents. Do not claim it independently confirms day 14 or any other date without the exhibit. Preserve correction lineage and the distinction between an index entry and the indexed evidence.
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
- Quoting the prior transcript page after an explicit correction.
- Treating an exhibit title or index as evidence of the exhibit's unseen contents.
Change the inputs and run it again
Add a later correction to page 5 only. Require page-4's day-14 source to remain T2 rather than assume every newest file replaces the entire transcript.
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 original fictional text. Exclude real proceedings, sealed files and case references; clear text and audio rights separately.
Use approved or fictional service records. Exclude customer and worker data, site access details and sensitive surveillance material.
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
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.
Compare the approach
micro1 review for experts, trade footage and company data
Compare the prerequisites for an approved recording project before planning a contribution.
Compare the approach
Sama review for computer-vision labeling and model evaluation
Review labeling scope for visual state and exception checks in staged footage.
Related industry examples
Industry guide · NAICS 561450
Credit bureau dispute and correction history data for AI
When a record is disputed, software needs to keep the challenged field, source and correction straight.
Industry guide · NAICS 561421
Telephone answering service callback handoff data for AI
An answering service can verify that a message was accepted into its queue even when the on-call team has yet to acknowledge it.
Industry guide · NAICS 561422
Contact center call escalation and quality rubric data for AI
Quality teams can test whether an evaluation follows the correct branch for a caller’s intent.