Short definition

The operation by which a source enters a knowledge base: it is read, its reusable elements are drawn from it, and the notes already present are updated, linked, or set in contradiction.

Full definition

In data processing, the word usually refers to loading documents or records into a system — a transport operation, which leaves the material as it was and touches nothing already stored.

In the articles on this blog, the word keeps the gesture of entry and changes what happens during it: ingestion is a work of writing on the existing. Its visible output is not the source filed somewhere, it is the neighbouring notes that have moved — an industrial standard or maintenance handbook whose entry into the base touches ten to fifteen already written notes.

The operation is governed by written rules — the "ingestion rules" — which define each type of note, its validity criteria, what must be detected and in what vocabulary it must be written. Those rules are part of the setup just as much as the notes are.

Usage in the field

Used to talk about a base's cost and quality: you calibrate ingestion rules, you measure an ingestion by the number of notes touched, you re-ingest a corpus after tightening the rules.

Synonyms and variants

"ingesting a source" for the act; "ingestion rules" for the body of instructions that governs it; "extraction pass" when the emphasis is on what is produced.

Not to be confused with

• indexing — building an index that allows a document to be found. It adds an entry without modifying the earlier ones; ingestion modifies the existing.
• Capture — the gesture of collecting a source or a thought on the spot, without processing. It builds the input stock; ingestion is what makes something of it.
• Distillation — the work of reducing material to what is essential in it. It describes an operation on a content; ingestion describes a source's entry into an already constituted whole.
• Persistence — the gesture by which a fragment is set aside so as to survive the exchange, without anything yet being decided about what it is worth. It precedes ingestion, which alone qualifies.

Examples

A source ingested that produces no new note and sharpens three glossary definitions: the operation succeeded, even though the base did not grow.

Ambiguities / debates

The word circulates between two worlds that do not load it the same way: on the data side, ingesting is neutral and automatic; on the knowledge side, ingesting implies judgement calls — what to keep, what to discard, what to connect. Using the term without lifting that ambiguity leads people to underestimate the work on rules that it covers.