1. Knowledge Organization Systems: An Overview
The term knowledge organization systems is intended to encompass all types of schemes for organizing information and promoting knowledge management1. Knowledge organization systems include classification schemes that organize materials at a general level (such as books on a shelf), subject headings that provide more detailed access, and authority files that control variant versions of key information (such as geographic names and personal names). They also include less-traditional schemes, such as semantic networks and ontologies. Because knowledge organization systems are mechanisms for organizing information, they are at the heart of every library, museum, and archive.
Knowledge organization systems are used to organize materials for the purpose of retrieval and to manage a collection. A KOS serves as a bridge between the user's information need and the material in the collection. With it, the user should be able to identify an object of interest without prior knowledge of its existence. Whether through browsing or direct searching, whether through themes on a Web page or a site search engine, the KOS guides the user through a discovery process. In addition, KOSs allow the organizers to answer questions regarding the scope of a collection and what is needed to round it out.
All digital libraries use one or more KOS. Just as in a physical library, the KOS in a digital library provides an overview of the content of the collection and supports retrieval. The scheme may be a traditional KOS relevant to the scope of the material and the expected audience for the digital library (such as the Dewey Decimal System or the INSPEC Thesaurus), a commercially developed scheme such as the Yahoo or Excite categories, or a locally developed scheme for a corporate intranet.
The decision of what knowledge organization system to use is central to the development of any digital library. The KOS must be applicable, either automatically or by human catalogers, to the resources included in the digital library. Once the material is included in the collection, the KOS must be meaningful to its users.
This section outlines the characteristics of KOSs, describes the common types, and discusses their origins and traditional uses.
Common Characteristics of Knowledge Organization Systems
It is often said that humans are inherent organizers. From an early age, children play sorting and matching games. We cope with our ever-changing world by comparing new objects or experiences with those with which we are familiar, identifying patterns and categorizing what is new into our existing frame of reference. The emphasis on developing comprehensive KOSs can be seen in the writings of our earliest philosophers, many of whom continue to influence our view of the world. For example, Aristotle's effort to categorize knowledge into groups (such as physics, politics, or psychology) is reflected in our language, our education, and our science. The original classification scheme of the Library of Congress, used between 1800 and 1814, was based on the philosophical works of Sir Francis Bacon and inherited from the English tradition. Beginning in 1814, the influence of Thomas Jefferson can be seen on the Library of Congress collection. Jefferson, who reclassified the library, reflected a more humanist philosophy (Lesk 1997).
There is no single knowledge classification scheme on which everyone agrees. Michael Lesk speculates that while a single KOS would be advantageous, it is unlikely that such a system will ever be developed. Culture may constrain the knowledge classification scheme so that what is meaningful to one culture is not necessarily meaningful to another (Lesk 1997). Therefore, we live in a world of multiple, variant ways to organize knowledge.
Despite their diversity, KOSs have the following common characteristics that are critical to their use in organizing digital libraries.
- The KOS imposes a particular view of the world on a collection and the items in it.
- The same entity can be characterized in different ways, depending on the KOS that is used.
- There must be sufficient commonality between the concept expressed in a KOS and the real-world object to which that concept refers that a knowledgeable person could apply the system with reasonable reliability. Likewise, a person seeking relevant material by using a KOS must be able to connect his or her concept with its representation in the system.
Types of Knowledge Organization Systems
A review of some typical knowledge organization systems shows their scope and applicability to a variety of digital library settings. While there are specific definitions for many of these KOSs in the computer science and information science literature, and even in standards documents, there is debate over these definitions. Terms are often used, particularly in the popular press and in the book trade, in nonstandard ways. Reflecting the scope of this practice, a recent National Information Standards Organization (NISO) workshop on electronic thesauri emphasized the need to improve the definitions of "terminology relating to terminology" (NISO 1999).
The descriptions given here provide an overview of possible systems for organizing digital libraries. The descriptions are based on characteristics such as structure and complexity, relationships among terms, and historical function. The list is not comprehensive; nor are the definitions of these terms contained in specific standards documents. They are grouped into three general categories: term lists, which emphasize lists of terms often with definitions; classifications and categories, which emphasize the creation of subject sets; and relationship lists, which emphasize the connections between terms and concepts.
Authority Files. Authority files are lists of terms that are used to control the variant names for an entity or the domain value for a particular field. Examples include names for countries, individuals, and organizations. Nonpreferred terms may be linked to the preferred versions. This type of KOS generally does not include a deep organization or complex structure. The presentation may be alphabetical or organized by a shallow classification scheme. A limited hierarchy may be applied to allow for simple navigation, particularly when the authority file is being accessed manually or is extremely large. Examples of authority files include the Library of Congress Name Authority File and the Getty Geographic Authority File.
Glossaries. A glossary is a list of terms, usually with definitions. The terms may be from a specific subject field or from a particular work. The terms are defined within a specific environment and rarely include variant meanings. Examples include the Environmental Protection Agency (EPA) Terms of the Environment.
Dictionaries. Dictionaries are alphabetical lists of words and their definitions. Variant senses are provided where applicable. Dictionaries are more general in scope than are glossaries. They may also provide information about the origin of a word, variants (by spelling and morphology), and multiple meanings across disciplines. While a dictionary may also provide synonyms and through the definitions, related words, there is no explicit hierarchical structure or attempt to group them by concept.
Gazetteers. A gazetteer is a list of place names. Traditional gazetteers have been published as books or have appeared as indexes to atlases. Each entry may also be identified by feature type, such as river, city, or school. An example is the U.S. Code of Geographic Names. Geospatially referenced gazetteers provide coordinates for locating the place on the earth's surface. The term gazetteer has several other meanings, including an announcement publication such as a patent or legal gazetteer. These gazetteers are often organized using classification schemes or subject categories.
Classifications and Categories
Subject Headings. This scheme type provides a set of controlled terms to represent the subjects of items in a collection. Subject heading lists can be extensive and cover a broad range of subjects; however, the subject heading list's structure is generally very shallow, with a limited hierarchical structure. In use, subject headings tend to be coordinated, with rules for how they can be joined to provide concepts that are more specific. Examples include the Medical Subject Headings (MeSH) and the Library of Congress Subject Headings (LCSH).
Classification Schemes, Taxonomies, and Categorization Schemes. These terms are often used interchangeably. Although there may be subtle differences from example to example, these types of KOSs all provide ways to separate entities into "buckets" or broad topic levels. Some examples provide a hierarchical arrangement of numeric or alphabetic notation to represent broad topics. These types of KOSs may not follow the rules for hierarchy required in the ANSI NISO Thesaurus Standard (Z39.19) (NISO 1998), and they lack the explicit relationships presented in a thesaurus. Examples of classification schemes include the Library of Congress Classification Schedules (an open, expandable system), the Dewey Decimal Classification (a closed system of 10 numeric sections with decimal extensions), and the Universal Decimal Classification (based on Dewey but extended to include facets, or particular aspects of a topic). Subject categories are often used to group thesaurus terms in broad topic sets that lie outside the hierarchical scheme of the thesaurus. Taxonomies are increasingly being used in object-oriented design and knowledge management systems to indicate any grouping of objects based on a particular characteristic.
Thesauri. Thesauri are based on concepts and they show relationships among terms. Relationships commonly expressed in a thesaurus include hierarchy, equivalence (synonymy), and association or relatedness. These relationships are generally represented by the notation BT (broader term), NT (narrower term), SY (synonym), and RT (associative or related term). Associative relationships may be more detailed in some schemes. For example, the Unified Medical Language System (UMLS) from the National Library of Medicine has defined more than 40 relationships, many of which are associative. Preferred terms for indexing and retrieval are identified. Entry terms (or nonpreferred terms) point to the preferred terms to be used for each concept.
There are standards for the development of monolingual thesauri (NISO 1998; ISO 1986) and multilingual thesauri (ISO 1985). In these standards, the definition of a thesaurus is fairly narrow. Standard relationships are assumed, as is the identification of preferred terms, and there are rules for creating relationships among terms. The definition of a thesaurus in these standards is often at variance with schemes that are traditionally called thesauri. Many thesauri do not follow all the rules of the standard but are still generally thought of as thesauri. Another type of thesaurus, such as the Roget's Thesaurus (with the addition of classification categories), represents only equivalence.
Many thesauri are large; they may include more than 50,000 terms. Most were developed for a specific discipline or a specific product or family of products. Examples include the Food and Agricultural Organization's Aquatic Sciences and Fisheries Thesaurus and the National Aeronautic and Space Administration (NASA) Thesaurus for aeronautics and aerospace-related topics.
Semantic Networks. With the advent of natural language processing, there have been significant developments in semantic networks. These KOSs structure concepts and terms not as hierarchies but as a network or a web. Concepts are thought of as nodes, and relationships branch out from them. The relationships generally go beyond the standard BT, NT, and RT. They may include specific whole-part, cause-effect, or parent-child relationships. The most noted semantic network is Princeton University's WordNet, which is now used in a variety of search engines.
Ontologies. Ontology is the newest label to be attached to some knowledge organization systems. The knowledge-management community is developing ontologies as specific concept models. They can represent complex relationships among objects, and include the rules and axioms missing from semantic networks. Ontologies that describe knowledge in a specific area are often connected with systems for data mining and knowledge management.
All of these examples of knowledge organization systems, which vary in complexity, structure, and function, can provide organization and increased access to digital libraries.
The Origin and Use of Knowledge Organization Systems
In the physical library, classification schemes such as Library of Congress (LC), Dewey Decimal System, and the Universal Decimal Classification reflect, among other things, the need to store a single item at a single location on a shelf. To provide multiple access points beyond the limits of a single physical location, subject headings are applied. Libraries use subject heading schemes such as LCSH, Sears, or other specialized schemes developed for specific content or specific collections. At the level of specific content, libraries have used authority files to control variant forms of personal, organizational, and geographic names.
However, KOSs can be found in settings other than libraries. An awareness of the KOSs available from alternative sources is valuable when considering the development of digital libraries for a specific audience.
Abstracting and Indexing Services
Abstracting and indexing (A&I) services developed as an outgrowth of traditional bibliographies and the explosion of journal literature. In the sciences, the development of A&I services was spurred by the post-World War I concerns about inadequate access to scientific information. In the 1950s, investment in A&I services was fueled by the Cold War and Sputnik. Abstracting and indexing services in the humanities, such as the Bibliography of the History of Art or the Modern Languages Association (MLA) Bibliography, generally took a different growth path than did their scientific and technical counterparts, but they also quickly became important resources for scholarship in the online environment. The scope of A&I services varies from broad discipline-oriented services (e.g., chemistry, architecture, biology, and physics) to narrowly defined aspects of the literature (e.g., peaceful uses of nuclear energy) and subdisciplines (e.g., aquatic sciences).
Special KOSs, such as thesauri and subject categories, were developed to support A&I services and their specific products and audiences. These organizations applied increasingly complex schemes to provide subject access to the literature in a variety of subjects. By the 1960s, A&I services were moving from the provision of print-only products to print and online services through large online vendors such as Dialog. Later, the products were distributed on CD-ROM and now, increasingly, on the Web. In many cases, the KOSs migrated from print to electronic media following the products they supported. While increased computing power, more sophisticated search engines, and more independent end-user searching have led to changes in some KOSs, most have retained their importance, even in the Web environment.
For many years, the KOSs related to A&I services were applied only by catalogers and indexers trained in using the KOS indexing for a particular product or products. The primary users of KOSs were librarians and other professional searchers. However, the proliferation of electronic data, the explosion of electronic publishing, and increasing concerns about the difficulty of locating information have led to a renewed interest in these KOSs for use not only by professionals but also by end users.
As publishers have migrated to electronic composition systems, they have become increasingly involved in the production of A&I products. Large journal publishers such as Academic Press and Elsevier have developed their own systems to provide bibliographic records linked to the full text of documents. As the content of online journals has grown, it has become necessary to move from systems that provide browsing by table of contents and journal issue to systems that support searching by both free text and by KOS. Electronic journals have resulted in additional KOSs, particularly classification and categorization schemes. For example, Elsevier's Web site has a subject categorization scheme to provide access to individual Web sites of its more than 2,000 titles.
Trade, Professional, and Governmental Organizations
A variety of authority files and classification schemes are used to support business and commerce. They range from the Standard Industrial Classification (SIC) code and the North American Industrial Classification System (NAICS), used in procurement and government statistics, to disease codes used to communicate patient illnesses and treatments among physicians, hospitals, and insurance companies. As more organizations develop Web sites, additional KOSs are being developed to support them.
Organizations are among the most prolific creators and users of KOSs. Developers of corporate intranets and knowledge management systems have discovered hundreds of specific classification schemes, glossaries, categorization schemes, and other vocabularies in use within organizations. Many of these are geared toward specific tasks and are, therefore, very narrow both in subject scope and target audience. However, for these audiences, they can also be rich sources of information.
For example, the Department of Energy (DOE) Environmental Management Science Program (EMSP) and the Office of Scientific and Technical Information are developing a digital library to support EMSP program managers. Program managers and researchers have developed "needs categories" and "science categories" to organize the Environmental Science Network (ESN). The categories are used primarily to support the process of grant submission and award; however, the ESN also uses them to provide access to related material from within DOE and from other distributed databases from the EPA, the Department of Defense, and NASA. Vocabulary is currently being organized around these categories for use with a Web mining tool that will provide highly relevant Web resources for project managers in specific areas.
Knowledge organization systems include a variety of schemes that organize, manage, and retrieve information. They range from authority files to classification schemes, thesauri, and ontologies. Libraries and other information management organizations have developed KOSs to organize and retrieve information. In addition to their primary function, which is to provide access to materials for a specific community or audience, KOSs can perform functions that further enhance the digital library.
1. The term knowledge organization systems as used in this report was coined by the Networked Knowledge Organization Systems Working Group at its initial meeting at the ACM Digital Libraries Õ98 Conference in Pittsburgh, Pennsylvania.