Knowledge Organisation

Overview

Knowledge organisation answers three questions for every item: what is it, what is it about, and where does it belong in relation to other items. Description answers what it is: title, author, edition, publisher, physical or digital form, and identifiers. Subject analysis answers what it is about: the topics, concepts, places, periods and forms that a searcher might use to seek it. Arrangement answers where it belongs: the shelf order, catalogue order, or result order that puts related items together and separates unrelated ones. A system that describes well but analyses subjects poorly will let users find known items but not explore topics. A system that analyses subjects well but describes poorly will retrieve relevant items that cannot be identified, distinguished or obtained. Good organisation balances all three.

Two traditions shape the field. The bibliographic tradition, developed for printed collections, emphasises controlled vocabularies, classification notation, catalogue codes and authority files, with human judgement at the centre. The retrieval tradition, developed for databases and web search, emphasises full-text indexing, ranking, relevance feedback and statistical measures, with algorithms at the centre. Modern practice combines them: controlled vocabulary improves precision, full text improves recall, authority control resolves name ambiguity, and linked metadata lets a record travel across systems. The examiner tests whether you can move between traditions: when to prefer a controlled term, when to add natural-language keywords, when a classification number helps browsing, and when a ranked keyword search helps discovery.

Universe of knowledge

The universe of knowledge is the totality of recorded subjects, concepts and their relationships, as distinct from any single library collection. It grows by specialisation, as broad subjects divide into narrower ones; by interdisciplinarity, as problems draw on several subjects at once; by fusion, as separate subjects merge into a new field; and by new phenomena, techniques and places that create subjects that did not exist before. A classification scheme must therefore be hospitable: it must provide room for new subjects without disturbing the whole arrangement, and it must allow a subject to be placed where searchers will look for it, even when it could logically belong in more than one place. The principle of helpful order states that arrangement should serve the searcher, grouping related subjects, placing general before specific where that aids browsing, and keeping the order predictable.

Subjects themselves are formed in several modes. A basic subject is a broad field such as physics or history. A compound subject combines a basic subject with facets such as place, time, form or viewpoint. A complex subject combines two or more basic subjects, or a basic subject with another idea in a relationship such as influence, comparison or application. Facet analysis breaks a compound subject into its component ideas, so that each facet can be represented systematically and combined by rule rather than by ad hoc invention. For example, a work on the treatment of a disease in children in a particular region can be analysed into subject, process, population and place facets, each contributing to its class number and subject headings. Modes of formation also include dissection, where a subject is divided into parts; denudation, where a broader subject is narrowed; lamination, where facets are layered onto a basic subject; loose assemblage, where two subjects are joined in a study relationship; and fusion or fission, where subjects merge or split over time. Recognising the mode helps a classifier choose notation that reflects structure rather than merely copying words from a title.

Knowledge organisation also involves mapping relationships: broader and narrower terms, related terms, synonyms and near-synonyms, and part-whole relationships. A thesaurus makes these relationships explicit so that a searcher who uses one term can be led to the preferred term, to broader context, to narrower specifics, and to related ideas. Without such mapping, retrieval depends on exact word matching, and a searcher who uses a synonym, a variant spelling, or a newer term will miss relevant material. The universe of knowledge is therefore not a fixed list of subjects but a changing network of concepts and relationships, and organisation systems must represent both the concepts and the links between them.

Classification

Classification assigns a subject a place in a systematic arrangement, expressed as a notation, and combined with author and work marks to form a call number that fixes shelf position. Its purposes are arrangement on shelves, arrangement in catalogues and bibliographies, subject retrieval, and statistical grouping of collections. A good scheme is comprehensive enough to cover the collection, hospitable to new subjects, expressive enough to show specificity, consistent in applying its rules, and practical in notation length and memorability. Notation may be pure, using one kind of symbol, or mixed, using letters and numbers. It may be hierarchical, showing broader-to-narrower relationships by extending the number, or faceted, building numbers by combining facet symbols according to a formula. Short notation is easier to shelve and recall, but long notation can express greater specificity and finer arrangement; every scheme trades these against each other.

The major general schemes illustrate different philosophies. The Dewey Decimal Classification divides knowledge into ten main classes, each divided by ten, giving a decimal hierarchy that is widely understood and relatively easy to learn, but that can force new or interdisciplinary subjects into an older structure and can produce long numbers for specificity. The Universal Decimal Classification extends decimal structure with auxiliary symbols for place, time, language, form and relationship, making it highly expressive for compound subjects and special collections, at the cost of complex notation. The Library of Congress Classification uses letters for main classes with detailed subclasses, developed for a very large research collection, giving depth in many fields but with uneven development and notation that is less mnemonic for beginners. The Colon Classification, developed by Ranganathan, is faceted: a subject is analysed into personality, matter, energy, space and time facets, and notation is synthesised by combining facet symbols, giving great hospitality and specificity for compound subjects, but requiring training in facet analysis. The Bibliographic Classification and other faceted or freely faceted approaches continue that tradition for special purposes. No scheme is best in the abstract. The right choice depends on collection size, user familiarity, subject emphasis, shelf arrangement needs, and compatibility with shared cataloguing and networks.

Practical classification follows a discipline: read the work, identify its subject and facets, determine its specific subject rather than a broad neighbour, check for rules on interdisciplinary, biographical or historical treatment, build or select the number, and verify that neighbouring numbers place it helpfully. Common errors include classifying by title words alone, choosing a number that is too broad, forcing a compound subject into a single facet, and ignoring facet order, which changes meaning: the facets must be cited in the order the scheme prescribes, or the number will arrange the work with the wrong neighbours. Classification is also a service decision: a number that is technically defensible but places a heavily used topic where its readers never browse has failed the principle of helpful order, and a local adaptation, documented and applied consistently, may be justified.

Cataloguing and bibliographic standards

Cataloguing creates a structured record describing a work and providing access points through which it can be found. Descriptive cataloguing records title, statement of responsibility, edition, publication, physical or digital description, series and notes, following rules that decide what to transcribe, what to supply in brackets, and how to handle variant titles and multiple authors. Choice of access points decides the main entry and added entries: which author, title, corporate body or uniform title serves as the primary heading, and which additional headings support searching by other names, titles and contributors. Subject cataloguing adds classification and subject headings or descriptors, linking the record to the subject system studied earlier. Authority control maintains authorised forms of names, titles and subjects, with references from variants, so that works by one author under different name forms, or on one subject under different terms, are gathered together rather than scattered.

Modern cataloguing rests on shared standards so that records can be created once and reused widely. Cataloguing codes have moved from earlier national codes toward internationally shared principles emphasising user tasks: find, identify, select and obtain a resource, and more recently explore relationships between works, expressions, manifestations and items. Bibliographic record formats encode records in machine-readable fields, subfields and indicators, allowing exchange between systems and display in catalogues and discovery layers. Standards for identifiers, such as international standard numbers for books and serials, support unique identification and linking, while authority files support consistent headings across institutions. Cooperative cataloguing, in which libraries share records and authority data, reduces duplication and improves consistency, but it requires agreement on standards, quality checks, and local adaptation for language, script and community needs.

The catalogue itself has changed from cards to online public access catalogues and discovery systems, but its functions remain: to show what a library holds, to help a user find a known item, to gather works by an author or on a subject, and to help select among editions and formats. A good catalogue record is therefore accurate, complete enough for identification and selection, consistent in headings, and linked to holdings and access: a record without location or licence information fails the obtain task even if description is perfect. Cataloguing judgement appears in edge cases: a work with no clear author, a title that changes across editions, a corporate author versus a personal author, a translation versus an original, and a serial that changes title. In each case the cataloguer applies principles — prefer the evidence in the work itself, record what aids identification, and maintain consistency — rather than inventing a rule for the single item.

Metadata, indexing and retrieval

Metadata is structured information about a resource, created to support discovery, identification, management, preservation and use. Descriptive metadata records title, creator, subject and description. Structural metadata records how parts relate, such as chapters in a book or files in a dataset. Administrative metadata records rights, provenance, technical properties and preservation actions. Schemas provide element sets for these purposes: a simple, widely shared element set supports interoperability across repositories, while richer schemas support detailed description for archives, museums or research data. Application profiles select and constrain elements for a particular community, balancing simplicity against specificity. Metadata quality matters as much as schema choice: incomplete, inconsistent or inaccurate metadata makes a well-preserved object undiscoverable, while clean, consistent metadata with controlled subjects and identifiers lets it be found, linked and reused.

Indexing is the process of representing the content of a document with terms, codes or other surrogates so that it can be retrieved. In precoordinate indexing, terms are combined at indexing time into a heading or class number, as in a subject heading string or a faceted class number, giving controlled, browsable order but requiring anticipation of search combinations. In postcoordinate indexing, single terms are assigned at indexing time and combined by the searcher at retrieval time with Boolean operators, giving flexibility for unforeseen combinations but depending on the searcher to combine well. Controlled vocabularies, including subject heading lists and thesauri, control synonyms, variant spellings and homonyms, and show broader, narrower and related terms. Natural-language indexing uses words from the text itself, improving recall for new terms and names, but suffering from synonym and ambiguity problems. Most modern systems combine both: controlled terms for precision and consistency, full-text words for recall and currency, with authority control to reconcile names and concepts.

Retrieval is judged by whether relevant items are found and irrelevant items are excluded. Recall is the proportion of relevant items in the collection that are retrieved. Precision is the proportion of retrieved items that are relevant. The two usually trade off: broadening a search raises recall but lowers precision, while narrowing raises precision but lowers recall. A search that retrieves everything has perfect recall and poor precision; a search that retrieves one perfect item has perfect precision and poor recall. Evaluation therefore reports both, and system design balances them for the task: a systematic literature search needs high recall, a quick factual search needs high precision, and a browsing search needs a helpful order even at moderate precision. Ranking, relevance feedback, query expansion with thesaurus terms, and faceted filtering all help users move along the trade-off deliberately rather than by accident. Understanding recall and precision turns searching from guessing keywords into designing a strategy with a measurable outcome.

Major classification schemes compared

Precoordinate versus postcoordinate indexing