document indexing
Enable an AI agent to recognise document indexing, assess whether assigned keywords, index terms or descriptors represent a document appropriately, and decide when those assignments need review or revision.
Bundle → Layer → Finding → Questions Filled
6 bundles · 11 layers · 18 findings · 28 questions
Retrieval purpose and boundaries Establishes what users need to find and which document units the indexing activity represents.
Retrieval tasks
Connects indexing choices to intended information-seeking tasks.
Intended retrieval outcome
Record whether indexing supports subject discovery, known-item lookup, passage retrieval, within-document navigation or a declared combination.
- What must a user or agent be able to find through this indexing activity? definition
- Which representative retrieval tasks will be used to judge whether the chosen indexing mode is suitable? measurement
Document and unit selection
Defines inclusion boundaries and the unit to which an indexing representation attaches.
Eligible content boundary
Record document versions, target granularity and treatment of attachments, tables, illustrations, notes and duplicate copies.
- Is each retrieval representation attached to a whole document, a structural section, a page or a generated passage? boundary
- Which content is deliberately excluded, and how are those exclusions distinguished from indexing failures? boundary
Content analysis and representation Describes how document content becomes a representation usable for retrieval.
Indexable content preparation
Records the content actually available to the indexing method and transformations that affect its meaning or boundaries.
Representation input
Identify source fields or passages and relevant extraction, language handling, normalisation and segmentation choices.
- Which source fields or passages were available for indexing, and which extraction failures reduced that input? provenance
- How do tokenisation, passage boundaries or language-specific transformations affect what can be retrieved? boundary
Representation assignment
Captures how terms, headings, fields or learned representations are selected and attached to content.
Assignment policy
Record whether representations are extracted, interpreted or generated, with their specificity and supporting content.
- What rule determines whether a concept receives an entry, including concepts discussed without being named explicitly? definition
- Which source content and indexing method support this assigned heading, term or machine representation? provenance
Index organisation and addressing Captures the structures through which retrieval representations lead to the appropriate content.
Vocabulary and entry relations
Records semantic and navigational relationships where the indexing mode exposes terms or headings.
Term identity and cross-references
Represent preferred terms, variants, ambiguous labels, subentries and cross-references when applicable.
- How are synonyms, spelling variants and identical labels with different meanings handled? definition
- Which cross-references or heading relationships must be checked for missing targets, misleading equivalence or circular navigation? action
Retrieval target addressing
Connects entries or machine representations to resolvable document identities and locations.
Locator integrity
Record the locator scheme and the document version against which addresses are valid.
- Does this representation resolve to a document identifier, page reference, anchor or passage identifier, and against which version? provenance
- What proportion of tested locators reaches the intended content rather than merely an existing destination? measurement
Indexing execution and maintenance Records responsibility for indexing decisions and keeps retrieval representations aligned with changing documents.
Assignment and review workflow
Identifies the people or methods that produced representations and the review required for their intended use.
Indexing decision trace
Associate indexing outputs with responsible indexers or processing configurations and any review decisions.
- Which indexer or automated configuration produced these entries, using which vocabulary or model version? provenance
- Which ambiguous assignments or processing failures require human review before acceptance? action
Change propagation
Determines how document and policy changes trigger indexing updates.
Affected-entry maintenance
Track entries affected by revision, deletion, relocation, access changes or replacement of an indexing method.
- Which changes require updating a locator, reassessing assigned concepts or regenerating the complete representation? action
- How are stale entries and representations of withdrawn or newly restricted content identified and removed from applicable retrieval paths? action
Indexing quality and retrieval evidence Assesses completeness, representational validity and usefulness under declared retrieval conditions.
Coverage and representation audit
Checks missing units, unsupported assignments and inconsistent treatment of comparable content.
Coverage and assignment defects
Separate processing coverage from the adequacy and support of assigned representations.
- How many eligible documents or content units lack a usable indexing representation? measurement
- Which sampled entries misrepresent the source, omit important concepts or apply the indexing policy inconsistently? measurement
Retrieval task evaluation
Tests the index through representative information needs while recording the surrounding retrieval configuration.
Fitness for retrieval
Connect evaluation results to specific indexing choices and distinguish indexing defects from downstream retrieval effects.
- Which query or navigation tasks, relevance judgements and success measures establish fitness for the intended users? measurement
- What evidence attributes a failed retrieval task to indexing rather than query interpretation, ranking, permissions or interface behaviour? boundary
Evidence and external alignment What the world already says about this thing, gathered so the model can be checked against it.
Reported evidence
Findings from the breadth pass, kept separate from the structural claims.
Check these first
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these check these first hold for the sense of document indexing this model covers, and on what evidence? provenance
Kinds and varieties
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these kinds and varieties hold for the sense of document indexing this model covers, and on what evidence? provenance
Standards and regulation
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these standards and regulation hold for the sense of document indexing this model covers, and on what evidence? provenance
Real-world use
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these real-world use hold for the sense of document indexing this model covers, and on what evidence? provenance
Typical measurements
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these typical measurements hold for the sense of document indexing this model covers, and on what evidence? provenance
Failure modes and hazards
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these failure modes and hazards hold for the sense of document indexing this model covers, and on what evidence? provenance
Regional variation
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these regional variation hold for the sense of document indexing this model covers, and on what evidence? provenance
Neighbouring kinds and how to tell them apart
Recalled without web access and unsourced; every item is a lead to verify.
- Which of these neighbouring kinds and how to tell them apart hold for the sense of document indexing this model covers, and on what evidence? provenance
Classifiers Filled
- Family
- Thing Registry
- Category
- Activities and processes
- Entry kind
- thing
- Plane
- ACT
- Domain
- ACT.ACT
What it is Filled
Document indexing is the activity of classifying a document by assigning or selecting keywords, index terms or descriptors that represent its subject content and support identification and retrieval.
In scope
- The retrieval purpose, intended users and document boundaries that guide term assignment are in scope.
- Assessment of document subjects and selection or extraction of representative terms are in scope.
- Keywords, index terms, subject headings and descriptors assigned to documents are in scope.
- Human, automated and hybrid term assignment, including review and correction, is in scope.
- Term coverage, retrieval evaluation and synchronisation with document changes are in scope.
Out of scope
- Document authorship, substantive editing and publication are outside this model.
- Scanning, transcription and optical character recognition are outside this model except for their effects on term assignment.
- General collection cataloguing and records management beyond subject term assignment are outside this model.
- Query processing, result ranking and search-interface design are outside this model except as requirements or evaluation contexts.
- Database performance indexes that accelerate storage access are outside this model.
- Vector indexing and full-text indexing without document classification by selected terms are outside this model.
- Back-of-book indexes and other navigation to locations within documents are outside this model.
Why it exists Filled
Enable an AI agent to recognise document indexing, assess whether assigned keywords, index terms or descriptors represent a document appropriately, and decide when those assignments need review or revision.
Distinguishing features Filled
- The activity assigns or maintains keywords, index terms or descriptors that classify a document by its subject content.
- The indexed unit is a document whose subject content is represented by the selected terms.
- Terms may come from a controlled vocabulary or be selected as free keywords.
- A summary can inform term selection, but summarisation alone does not establish document indexing.
- Automated methods belong here when they select or assign classification terms to documents.
- Vector representations, hyperlinks and within-document locators alone do not establish document indexing in this sense.
What robots and AI may and may not do Filled
Must not
- Index documents the agent is not authorized to access.
- Expose confidential or personal content through search results.
- Drop documents from an index in a way that hides them from legal discovery.
- Apply biased or offensive subject terms to people or groups.
Only with a human decision
- Making a private collection searchable to a wider audience requires human review.
May
- Classify documents the agent is authorised to read using keywords, index terms or descriptors.
- Assign subject descriptors following a stated vocabulary or standard.
Moral aspects Filled
- Indexing decides what can be found; omissions and labels shape knowledge.
- Search can expose private data.
Who is affected
- People named in documents
- Users of the collection
Owners Filled
Steward
The owner of the collection.
Links to other meta-models Filled
parent
- Q11977025 - registry parent class
- Q2210408 - registry parent class
What else AI and robots need to interact with it Filled
Identity and identifiers required Filled
- Vercy registry: vr.tr.document-indexing
- Wikidata: Q1167863 (https://www.wikidata.org/wiki/Q1167863)
Direct properties not applicable Not applicable
- Indexing mode: Subject indexing; within-document indexing; full-text indexing; vector indexing; hybrid; other declared mode
- Retrieval target granularity: Document; section; page; paragraph; passage; other declared unit
- Indexed document version: Link between an indexing result and the document version or content snapshot it represents
- Representation basis: Controlled vocabulary; free terms; extracted tokens; structured fields; learned representations; combination
- Eligible-unit coverage: Percentage of declared eligible document units successfully indexed, with denominator and exclusions recorded
- Index freshness lag: Elapsed time between a relevant document change and its incorporation into the index
- Review state: Unreviewed; under review; accepted for declared use; correction required; superseded
- Retrieval evaluation result: Metric name and value with query set, relevance judgements, index version and retrieval configuration
Plane ACT: no invented physical properties.
Recognition not applicable Not applicable
Not applicable
Document indexing is an activity whose identity cannot be established from visual appearance alone.
Capabilities and actions required Filled
- Declare which documents and content units should be indexed for a specified retrieval purpose.
- Generate or assign retrieval representations using a declared indexing policy.
- Inspect index entries against source passages and correct unsupported or misleading representations.
- Validate locators, cross-references and coverage before accepting an indexing result.
- Compare indexing alternatives against representative retrieval tasks.
- Identify and update or remove entries affected by document revision, withdrawal or access changes.
Hazards and failure modes required Filled
- Incorrect or incomplete indexing can hide relevant documents or retrieve misleading material.
- Biased descriptors can misrepresent people or subjects and systematically distort discovery.
- Searchable metadata can expose sensitive information even when the underlying document has restricted access.
- Outdated index entries can direct users or agents to superseded or unavailable records.
Standards and interfaces required Filled
- ISO 5963, Documentation - Methods for examining documents, determining their subjects, and selecting indexing terms, International Organization for Standardization.
Context of use required Filled
- Document indexing is used in libraries, archives, search systems and organizational records management to make documents discoverable.
- It connects document content with search terms or controlled descriptors so users and agents can retrieve relevant material.
Sources Filled
- Wikidata item Q1167863: document indexing - identity and sense of the item
- Wikipedia: Subject indexing - general description of the item
Open questions
- Does the registry intend document indexing to encompass subject indexing, back-of-book indexing and automated search indexing, or does an existing registered or world-model concept own part of that scope?
- Which authoritative indexing traditions and sources should establish the definitions and applicability of specificity, exhaustivity and consistency for each included mode?
- How should the model assess evidence for learned retrieval representations whose relationship to individual document concepts is not directly inspectable?
- Which retrieval tasks, relevance judgements and acceptance thresholds are appropriate for the intended document collections and users?
- Where should responsibility lie for removing sensitive content from retrieval representations, including snippets and embeddings, when source-document access changes?
Machine files
Provenance
thing registry research (pass 2) · unreviewed
Built from: models/things/publications/thing-q1167863/spec.json