NERVE is live across local, private and approved-cloud deployments. Book an evaluation
Audience
Teams who need extraction to improve without invalidating work already done.
Source families
All supported families
The workflow
01
Input
An approved parser or schema update, and a defined scope of datasets to reprocess.
02
Acquisition and preservation
Nothing is re-acquired. Reprocessing reads the retained originals that were accepted the first time, which is why the operation is repeatable.
03
Transformation
The campaign is bounded and versioned. New canonical outputs record the provenance that produced them, including which parser and which schema version was used.
04
Discovery and retrieval
New outputs coexist with prior outputs. A citation created before the campaign still resolves, and comparison warnings highlight where the new extraction disagrees with the old one.
05
Governance
Reprocessing does not widen access. Handling, privacy and authority metadata are carried forward, and a record cannot become more visible because it was reprocessed.
When it goes wrong
Failure variant
The new run fails partway. Prior durable state is left intact and visible, the campaign reports where it stopped, and the earlier outputs continue to serve search and citation.
What this does not claim
Improving extraction changes the derived layer. The original accepted bytes and the canonical records already published remain exactly as they were.
Evaluation measures
Reproducibility of a campaign against the same originals
Comparison warnings between old and new outputs
Old-reference resolution after restart and reindex
Bounded campaign behaviour under load
Effort to explain what changed and why