Known-item search¶

Known-item search is retrieval undertaken with a particular target item already in mind and identifiable by attributes such as author or title.

Version: v1 · 2026-09-28 · History
Domain-specific #: 7682
Domain group: Professional & Organizational Practice
Origin domain: Library & Information Science
Subdomain: Information Retrieval → Library & Information Science

Core Idea¶

A known-item search is an information-retrieval task in which the searcher already has a particular target item in mind and seeks to locate its record or accessible instance.^[1] The target's identity is prior to the query, even when the searcher remembers only some identifying attributes.^[2]

In a library catalog, an author, title, publication detail, or combination of partial bibliographic cues can be used to recover the intended book, article, or other work.^[3] On the web or in another online collection, the cues may instead include a distinctive phrase, filename, creator, site, date, or remembered content.^[4] Query reformulation resolves incomplete or inaccurate recollection against the system's available fields and ranking.^[5]

Success is item-specific: the returned result must be the particular intended object, not merely a document relevant to the same topic.^[6] This makes known-item performance sensitive to name variation, metadata quality, spelling, indexing, entity disambiguation, and whether the system recognizes partial identifiers.^[7]

Known-item search is distinct from exploratory search, in which the searcher is learning the domain, refining an uncertain goal, or discovering what kinds of results exist.^[8] A person may alternate between the modes—exploration can reveal an item later sought directly—but topical relevance alone does not convert a search into known-item retrieval.^[9] The constitutive test is whether a determinate target was already intended before the retrieval process began.^[10]

Structural Signature¶

Sig role-phrases:

the searcher — the person undertaking retrieval with a particular item already intended before the immediate search begins.
the prior target identity — the determinate book, article, page, file, or other information object whose recovery defines success.
the recalled cues — partial or exact author, title, phrase, date, creator, site, filename, or content attributes remembered about the target.
the searchable collection — the catalog, archive, web index, or file system that may contain a record or accessible instance of the item.
the indexed representations — metadata fields, names, text, and identifiers against which the recalled cues can match.
the query and reformulation loop — cues are added, removed, corrected, or varied to resolve imperfect recollection against the available representation.
the candidate result set — records returned by matching and ranking, including relevant neighbors that are not the intended object.
the item-specific success test — retrieval succeeds only when the returned record or instance is the particular prior target.
the failure branches — cue error, name variation, ambiguous identity, weak metadata, indexing gaps, ranking, or access can each prevent recovery.
the exploratory boundary — when no determinate target precedes the search and any useful topical result may satisfy the need, the task is exploratory rather than known-item retrieval.

What It Is Not¶

Not topical search for any relevant item. Success requires recovery of the particular object intended before the query, not merely a useful result about the same subject.
Not exploratory search. When the searcher is learning what exists or deciding what would satisfy the need, the goal is still being formed rather than directed at a known item.
Not restricted to exact identifiers. A search can remain known-item retrieval when the author, title, phrase, date, site, or other cues are incomplete, variant, or partly misremembered.
Not defined by one interface or collection. Catalogs, archives, web indexes, and file systems can all support the task when they expose representations against which cues for the prior target can be matched.
Not successful merely because the target appears somewhere in the result set. Ranking, disambiguation, and access matter if the searcher cannot identify and reach the intended record or instance.
Not proof that a failed item is absent. Cue error, name variation, weak metadata, indexing gaps, ranking, collection coverage, or access failure can prevent recovery even when the object exists.^[11]

Scope of Application¶

Known-item search applies to retrieval episodes in which a searcher has a determinate information object in mind before querying and success requires recovering that particular record or accessible instance, even when its identifying cues are incomplete or inaccurate.

Library catalogs. Searches by known author, title, edition detail, or partial bibliographic combination are the concept's originating habitat.^[12]
Bibliographic databases. Article, book, proceedings, and report records can be recovered from remembered creator, title, venue, date, or citation fragments.
Web search. A user may seek one previously encountered page, document, image, or site using remembered phrases, creator names, URLs, or content details.^[13]
Digital libraries and archives. Known manuscripts, recordings, photographs, datasets, or other collection objects are sought through descriptive metadata and indexed content.
Institutional repositories. A particular thesis, report, preprint, or deposited object can be targeted by partial author, title, date, department, or identifier cues.
Personal and shared file search. A determinate file can be recovered through filename, path, creator, date, or remembered text when the storage system exposes those representations.
Catalog and search-interface design. User-centered features for fielded search, name variation, spelling correction, query reformulation, and disambiguation can be evaluated against item-specific tasks.
Information-retrieval evaluation. Test collections and user studies can measure whether the intended item is identifiable and reachable, rather than counting any topically relevant result as success.^[14]
Metadata-quality diagnosis. Failed recovery can expose absent fields, variant names, incomplete descriptions, entity conflation, or weak authority control for a known target.
Indexing and ranking diagnosis. When a target exists but remains unrecovered, the episode can isolate representation, collection coverage, candidate ranking, or access as the failure site.
Mixed search sessions. The identity applies only to those phases after a particular target is fixed; earlier domain learning or open-ended result discovery remains exploratory search.

Clarity¶

Known-item search makes success item-specific rather than merely topical. A result can be highly relevant to the remembered subject and still fail if it is not the particular book, article, page, or file intended before the query began. Partial author, title, phrase, date, site, or content cues are evidence used to recover that identity; they need not themselves be exact or complete.

The label also separates retrieval failure from goal formation. In exploratory search, the user is learning what exists or refining an uncertain need; in known-item search, reformulation resolves imperfect recollection against metadata and indexes. A session may alternate between the two, but the immediate question is: was there a determinate target before this search, and which identifying cues distinguish it from merely similar results? This exposes the roles of name variation, misspelling, metadata quality, and entity disambiguation.

Manages Complexity¶

A catalog or web index may contain millions of topically related records, while the searcher may remember only fragments of one intended item. Known-item search reduces that retrieval problem to a target identity, a set of recalled discriminating cues, the fields or representations against which those cues can match, and an item-specific success condition. A searcher or evaluator can then read off whether reformulation is resolving an author, title, phrase, creator, date, or site variation—and whether the intended record was recovered rather than merely a relevant neighbor.

Library-catalog and web-search branches use different metadata and ranking signals, and cues can range from exact identifiers to incomplete or misspelled recollections. The reduction stops before target identity and retrieval performance become trivial. Ambiguous editions or versions, uncertain memory, name variation, poor metadata, indexing gaps, access failure, and ranking behavior still require investigation. A failed session may reflect the cue, representation, collection, or interface; the task label alone does not locate the fault or turn exploratory goal formation into known-item retrieval.

Abstract Reasoning¶

Diagnostic inference moves from a determinate intended item, the searcher's recalled cues, and the returned records to the kind and likely location of a retrieval failure. Results about the right subject but not the intended item show failure under the item-specific criterion; a recognizable target absent under one spelling, title form, or creator name directs attention toward cue variation, metadata, or indexing rather than toward topical relevance alone.

Interventionist inference moves from changing the query's identifying cues to a prediction about the candidate records exposed by the retrieval system. Adding a discriminating author, title phrase, date, or site restricts matches that lack that attribute; replacing an inaccurate or variant form can expose a differently represented record; removing a cue can recover an item that the over-specified query excluded. These predictions remain conditional on the collection containing and indexing the target and on the interface matching the chosen representation.

Boundary inference moves from the searcher's goal state before an immediate search to its classification. If a particular item was already intended, even under partial recollection, the episode is known-item search; if the user is still learning what exists or deciding what would satisfy the need, it is exploratory. When a session changes from discovery of an item to later attempts to retrieve that same item, the classification changes with the goal state rather than with the topic or interface.

Knowledge Transfer¶

Within information retrieval, known-item search transfers literally across library catalogs, archives, websites, and file systems when the searcher has a particular target before querying. The cargo that carries intact is prior target identity, recalled author/title/phrase/date/site or content cues, searchable fields and representations, query reformulation, candidate records, and item-specific success. Diagnostics transfer by distinguishing cue mismatch, metadata or indexing failure, access failure, and a result that is merely topically relevant.

Beyond retrieval systems, the honest case is (B) shared target-recovery mechanism. Locating a known person or object from partial clues has the same search shape, but the home-bound cargo is an indexed information collection, query interface, records, and document identity. Exploratory or topical search is not a weak instance; it has a different goal because the target is not fixed in advance. The stopping boundary is prior identity: if success can be satisfied by any useful item about a topic, known-item metrics and reformulation diagnostics do not transfer.

Examples¶

Canonical¶

A researcher wants the particular 1995 article by Barbara Wildemuth and Ann O'Neill, “The ‘Known’ in Known-Item Searches,” but remembers only the two authors and the distinctive phrase “known in known-item.”^[15] In a library catalog, the first query returns records about known-item searching as well as the desired article. Adding the journal title College & Research Libraries and year removes topical neighbors and exposes the intended record.^[16] The task succeeds only when that article is identified and reached; a different useful article about known-item retrieval would still be failure because the researcher's target preceded the query.^[17]

Mapped back: the researcher is the searcher, and the 1995 article is the prior target identity. Author names, phrase, journal, and year are the recalled cues; the library catalog is the searchable collection, whose author, title, journal, and date fields are the indexed representations. Adding fields performs the query and reformulation loop, narrows the candidate result set, and meets the item-specific success test only when the intended article is recovered.

Applied / In Practice¶

A web user is trying to relocate the SIGIR 2003 paper “Combining Document Representations for Known-Item Search” after previously encountering it.^[18] The user initially remembers only “document representations” and “known item,” so a broad query surfaces many pages about retrieval methods. Adding the remembered conference name and the authors Ogilvie and Callan distinguishes the paper's record from same-topic results. If the user instead begins with no particular paper in mind and is simply looking for any account of document representation, the apparently similar session is exploratory; discovering this paper may create a later known-item target, but it was not one at the outset.

Mapped back: the previously encountered paper provides the prior target identity, and its phrase, conference, and authors provide the recalled cues used by the searcher. A web index serves as the searchable collection, with page text and bibliographic metadata as the indexed representations. Refinement changes the candidate result set until the item-specific success test is met. The alternate no-target session crosses the exploratory boundary, while failure under a name or indexing mismatch would instantiate the failure branches rather than prove absence.

Structural Tensions¶

T1: Prior target determinacy versus imperfect recollection. Known-item search requires a particular item to be intended before the immediate query, yet the cues available to the searcher may be fragmentary, variant, or wrong. The target can be determinate even when its description is not. Diagnostic: ask whether one returned object would uniquely satisfy the pre-existing goal, then separate uncertainty about its identity from uncertainty about its attributes.

T2: Discriminating specificity versus query overconstraint. Adding author, title, date, phrase, or site cues can remove topical neighbors, but one inaccurate cue can also exclude the intended record. Reformulation must sharpen identity without turning fallible memory into a mandatory filter. Diagnostic: vary one cue at a time and observe whether the candidate set narrows toward or unexpectedly loses the target.

T3: Item persistence versus representation dependence. The sought work may exist while catalog metadata, indexed text, name forms, or identifiers fail to expose it under the remembered cues. Known-item failure therefore need not imply item absence. Diagnostic: test alternate representations and collection coverage before concluding that the target is unavailable.

T4: Topical relevance versus item-specific success. A retrieval system can return excellent material about the right subject and still fail the task if the particular intended object is missing or unrecognized. This strict criterion preserves the task's identity but can make conventional relevance measures misleading. Diagnostic: verify the returned record against the prior target rather than scoring success from subject similarity alone.

T5: Result-set presence versus usable recovery. A target may technically appear in a long or poorly ranked result set while remaining effectively unrecovered to the searcher, and a visible record may still be inaccessible. Counting presence alone favors system recall over the user's actual goal. Diagnostic: record whether the searcher can identify and reach the intended instance, not merely whether it exists somewhere among candidates.

T6: Mode continuity versus session transitions. A session can begin exploratorily, form a target through discovery, and then become known-item search, or return to exploration after a failed retrieval. Treating the whole session as one mode hides the changing goal state. Diagnostic: classify each search episode by whether a determinate target existed immediately before that query.

T7: Search-and-Retrieval reduction versus known-item autonomy. The exact parent Prime Search and Retrieval strictly subsumes the task: every qualifying known-item search begins from an information need, traverses a search space through cues and matching criteria, produces candidates, and ends in access or failure. The task remains in situ because one determinate information-object identity precedes the query and provides an item-specific success test despite imperfect recollection. Reduction gains portable need–space–query–candidate–retrieval structure but erases the prior-target condition; complete autonomy hides the retrieval loop. Diagnostic: if the pre-query item identity is removed while indexed traversal and candidate recovery remain, Search and Retrieval survives but Known-Item Search does not.

Structural–Framed Character¶

Known-Item Search is mixed. Its evaluative_weight is low-medium: success is judged against a determinate prior target rather than general topical usefulness, but that criterion does not rank the target itself. Its human_practice_bound character is medium-high because the search mode depends on an agent already intending one information object and carrying incomplete remembered cues; an identical query can be exploratory when that intention is absent. Its institutional_origin is medium: library and information-retrieval practice stabilized the distinction and its evaluation methods, although catalogs, web indexes, and file systems can realize it without one institution. Its vocab_travels judgment is medium: query, target, index, matching, ranking, and retrieval move readily among information systems, while known-item retains its technical task distinction. Its import_vs_recognize profile is mixed: the retrieval chain is recognizable in the interaction, but classifying it as known-item requires importing the searcher's pre-query goal state and an item-specific success rule.

The smallest positively reviewed portable skeleton is Search and Retrieval: an information need is pursued through a represented search space, matching criteria, traversal, candidate results, and access or failure. Known-Item Search adds the domain-bound condition that one determinate information-object identity precedes the query and alone fixes success. Remove that prior target and the general retrieval structure remains as topical or exploratory search; remove the retrieval loop and remembered identity never becomes a known-item search. The cross-domain reach belongs to that Prime.

Its character: a formally tractable retrieval pattern whose identity remains partly framed by human intention, memory, and an item-specific criterion of success.

Structural Core vs. Domain Accent¶

Known-Item Search is a domain-specific specialization of the Prime Search and Retrieval: an information need is pursued through a represented search space, matching and traversal produce candidates, and access to a result completes the operation. Its differentia is that one determinate target identity exists before the query and alone fixes success.

What is skeletal (could lift toward a cross-domain prime). Search and Retrieval supplies a sought object or information need, a search space and its representation, matching criteria, a traversal or query operation, candidate results, and a recovery or access outcome with explicit failure modes. That signature recurs in at least three unrelated domains—for example, a library patron searches a catalog, a developer searches a file system, and an archival scholar searches indexed records. Known-item search fills every role with remembered attributes, a catalog or index, reformulation, ranked records, and access to the intended item.

What is domain-bound. Information-retrieval practice supplies a searcher who already intends a particular book, article, page, file, or other object; partial author, title, phrase, date, filename, site, or creator cues; and an item-specific success test. Metadata variation, spelling, entity ambiguity, indexing gaps, ranking, collection coverage, and access supply distinct failure branches. The exploratory boundary is constitutive: the same query can be exploratory when no determinate target precedes it. Remove this pre-query intention and general retrieval remains, but known-item search does not.

Why this does not clear the prime bar. Stripping target-memory and catalog vocabulary leaves Search and Retrieval's need–space–match–traverse–recover structure, already complete across unrelated domains. Conversely, retain a remembered item but remove the operational search and candidate-recovery loop, and memory of the object alone is not a search. Both removal directions establish strict subsumption: the Prime remains autonomous, while the child requires the agent's prior target identity and the stronger criterion that topical relevance cannot substitute for recovery of that exact item.

This entry is a kind of Search and Retrieval.

Strictly instantiates — Search and Retrieval (Search and Retrieval). The remembered item supplies a determinate information need; the catalog, archive, web index, or file system supplies the search space; recalled author, title, phrase, date, or other cues supply the matching criteria; indexes and query reformulation perform traversal; ranked records form the candidate set; and access to the intended record is the retrieval outcome. Removing that search-to-retrieval chain leaves only memory of an item. Preserving it without a target fixed before the query and an item-specific success test leaves general or exploratory search, which is exactly the domain-specific residual.

Evaluation is declined as the parent. Comparing returned candidates with the prior target is a contained judgment, but metadata, indexing, ranking, collection coverage, or access can make the retrieval task fail before such a verdict is available.

Relationships to Other Abstractions¶

Current abstraction Known-item search Domain-specific

Parents (1) — more general patterns this builds on

Known-item search is a kind of Search and Retrieval Prime

The remembered item supplies a determinate information need; the catalog, archive, web index, or file system supplies the search space; recalled author, title, phrase, date, or other cues supply the matching criteria; indexes and query reformulation perform traversal; ranked records form the candidate set; and access to the intended record is the retrieval outcome.

Hierarchy paths (4) — routes to 3 parentless roots

Known-item search → Search and Retrieval → Problem Space → Representation → Abstraction

Show alternative paths (3)

Neighborhood in Abstraction Space¶

Known-item search sits in a sparse region of the domain-specific corpus (73^rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Memory Encoding & Retrieval Effects (20 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Exploratory search. Exploratory search develops an uncertain goal or learns what a domain contains, whereas known-item search begins with one determinate target already intended. Tell: ask whether more than one previously unknown useful result could satisfy the immediate need.
Topical search. Topical search seeks relevant material about a subject; known-item success requires recovery of the particular intended object even when another result is highly relevant. Tell: compare each result with the pre-query target rather than with the topic alone.
Navigational search. Navigational search commonly aims to reach a particular site or page, but its task class is defined by a destination-oriented web intent rather than the broader recovery of any known information object across collections. Tell: determine whether the target is specifically a destination to visit or an item to identify and recover from a represented collection.
Exact-identifier lookup. Identifier lookup retrieves an object from a complete key such as a call number or DOI; known-item search also covers partial, variant, or misremembered author, title, phrase, date, and content cues. Tell: check whether matching is direct on a complete identifier or requires resolving incomplete recollection.
Relevance retrieval. A relevance-ranked system can return strong subject matches without returning the intended item, so ranking quality alone does not establish known-item success. Tell: require identification and access of the prior target, not merely a high-scoring neighbor.
Recommendation. Recommendation proposes items likely to suit a user, whereas known-item search recovers an object whose identity predates the query. Tell: ask whether the system is selecting an acceptable item or locating the one already wanted.
Browsing a collection. Browsing exposes records through categories, links, or ordered displays without necessarily fixing a target in advance. Tell: if the user is inspecting what exists rather than testing candidates against one prior identity, the activity is not yet known-item search.
Proof of absence. Failure to recover a known item can arise from wrong cues, name variation, weak metadata, indexing, ranking, collection coverage, or access. Tell: distinguish failure of the search path from evidence that the object does not exist.

References¶

[1] Jin Ha Lee, Allen Renear, and Linda C. Smith, “Known-Item Search: Variations on a Concept” (2006) (source). registry ↩ Show verification details

[2] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[3] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[4] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[5] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[6] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[7] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[8] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[9] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[10] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[11] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[12] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[13] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[14] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[15] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[16] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[17] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[18] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩