This section emphasizes the ability of knowledge organization systems to link digital library resources to other related resources. The basis for this linking is the identification of information within a digital resource that can be extracted and used to search and locate information within a KOS. The KOS may then be used to expand codes to more explanatory full text, to provide more descriptive records, or to link entity names to resources of physical specimens.
Expanding Codes to Full Text
Practitioners of a discipline use coding schemes to facilitate communication within that discipline. It is often helpful to connect these coding schemes to the full names for which the code stands. The examples provided here include links between databank registration codes and the biological sequence data, and between industrial codes and the full name that the code represents.
Linking Sequence Numbers to Biosequence Databanks
The lengthy biochemical and genetic sequences that molecular biologists, biotechnologists, and geneticists identify each day are kept in databanks. Several databanks have been developed, for example, to cover protein sequences, nucleotides, and cell lines. One of the largest databanks contains information on the mapping of the human genome. As molecular biologists began to discover these sequences, they reported them in scientific journals. Difficulties in composing, proofreading, and printing the text soon arose. Through an ad hoc standards process, major biomedical publishers agreed to require the inclusion of codes or databank numbers for these sequences in articles when they are published. In addition, the sequence itself must be registered in a databank before the paper can be published.
Some of the most frequently referenced databanks are listed on the Web site of the National Center for Biotechnology Information. They include GenBank and the Research Collaboratory for Structural Bioinformatics Protein Data Bank. Each sequence number is different, but all begin with a persistent code identifying the databank.
How can the link be made between the literature and the databank? Through a search profile, a text analysis program, or keyword indexing, the text can be analyzed and the sequence databank numbers identified. An active link can be embedded. The active link consists of a search strategy (possibly written as a CGI script) to locate that sequence number in the databank where the actual sequence is stored. When the user clicks on the active link, the script is generated and launched from the user’s browser. The Web-enabled database is searched, and the sequence record is returned to the user. Depending on the services provided by the databank site, the user can analyze the sequence using a number of tools provided by the databank or download the sequence for local manipulation.
This type of connection exists between the National Library of Medicine’s (NLM) search service, PubMed, and GenBank at the National Center for Biotechnology Information. If a search in PubMed yields records that have GenBank numbers, the user can automatically search and display the sequence records from GenBank.
Linking Individual Industrial Codes to the Full Scheme
In business, classification schemes serve to communicate important facts about a company or product. These codes are generally controlled by a government, professional, trade, or international standards organization. They often serve as shorthand for users interested in material in a particular area of industry or a specific business sector.
Perhaps the most familiar scheme is the SIC code, which was last updated in 1987. The SIC codes have been used by the U.S. government, economists, financial markets, regulators, and procurement offices to identify manufacturing, agriculture, and service sectors of the economy. In 1997, a new scheme was approved for use within the United States. The North American Industrial Classification System was developed with Canada and Mexico as a means of providing an agreed-upon scheme for the collection, reporting, and analysis of information about the economy by sector, both within and across borders. Information about NAICS is available from the Web site of the U.S. Census Bureau (see references for address).
The digital library can provide related information by using the authority files for the coding schemes as a linked authority file. If a company or economic sector mentioned in the digital library’s collection can be linked to an SIC or NAICS code, the code can be searched against the official tables of definitions maintained by the U.S. Census Bureau. These files provide definitions of the codes and place each code in the classification scheme with other economic sectors.
The digital library’s content can be further enhanced by making a link between the SIC and NAICS codes. If the digital library resource has the SIC code, it can be extracted and searched against the Census Bureau’s 1997 NAICS and 1987 SIC Correspondence Tables. The table returns the corresponding code from the alternate scheme.
Linking to Descriptive Records
Linking the name of an entity, such as a personal name, organization, or location, to additional information about that entity was one of the first uses of hyperlinking. Knowledge organization systems such as dictionaries, glossaries, and classification schemes can be used to link the entities in one resource to richer descriptions of that entity in another resource. This is particularly helpful for users who are new to a topic and in cases where the additional information can make the user’s task more efficient.
The examples that follow are from three disciplines. The first example links organism names to records that not only describe the species more fully but also put it in the context of the overall classification scheme for living organisms. The second example links chemical names to descriptive records and molecular structures. In the third example, proper names are linked to the biographies for the person.
Linking Organism Names to Taxonomic Records
Genus-species names are the Latin names for organisms e.g., plants, animals, and microorganisms. Taxonomists, who study and classify living organisms, create records for each of these organisms. Generally, these records are linked relationally to the other organisms in a hierarchy. Beyond the organism name and the information that it and its placement in the hierarchy convey, taxonomic records use other elements to describe the organism. These may include distribution patterns, the authority for naming and classification, and the date the organism was identified. Scientists base the information on specimens that are retained because they serve as the physical evidence of the description. Natural history museums, private collections, and individual scientists number, or code, the specimens in their collections. Sometimes specimens are supported by photographs or line drawings, which may be digitized.
By using a taxonomic authority file as an intermediate authority file, one can link a text or an image file containing a name or picture of an organism to additional related information. By automatically processing the text or embedding a link from the organism name in the text or from the image to the taxonomic authority record, one can extend the knowledge conveyed by the text. The text can include the descriptive and historical information in the taxonomic record and, ultimately, link to a photograph, a drawing, or appropriate video or audio segments.
Because of the ambiguity in organism names, many examples of this type are now created manually. However, depending on the extent of the files involved, the ambiguity of the Latin and common names for organisms can be overcome. An example of a taxonomic intermediate file is the Integrated Taxonomic Information System (ITIS). ITIS is a partnership of U.S., Canadian, and Mexican government agencies, private organizations, and taxonomic specialists cooperating to develop an online, scientifically credible list of biological names of North American plants and animals. It is used by many U.S. government agencies for consistent naming of plants and animals for regulatory and monitoring purposes. To link textual material in a digital library to the ITIS record, the organism name can be identified manually or automatically in the text and submitted as a query to the ITIS database. When a match is found, ITIS presents the ITIS record, which provides essential information about the organism. The information includessynonymous names, including some common names, and an indication of the placement of the organism in the larger taxonomic classification scheme.
Linking Chemical Names to Molecular Structures
The unique identification for a chemical substance is not its name but its molecular structure. However, chemical names are commonly used in research documents, project plans, catalogs, and directories, all of which may be resources in a digital library. There are competing systems of nomenclature (i.e., that of the Chemical Abstracts Service [CAS] and of the International Union of Pure and Applied Chemistry) as well as common and commercial synonyms.
The ambiguity is resolved by providing links between the chemical names in the text and the molecular structure. This is done through a chemical registry number or code that is connected to a particular chemical name (using certain nomenclature standards) and an authority record that provides additional information about the chemical. This information includes the chemical’s synonyms and some of its chemical and physical properties. Most important in today’s research environment is the link from this authority file to a chemical structure file. Structure files, used with the appropriate software, graphically depict the molecular structure. This sophisticated software allows for three-dimensional visualization, rotation, and substitution of the chemical bonds.
An example of the use of the chemical registry number to link chemical names with molecular structures can be seen in the work of BIOSIS, the world’s largest not-for-profit producer of biological and biomedical databases. In 1993, BIOSIS began processing its bibliographic citations (titles and keywords) to automatically identify chemical names (Hodge, Nelson, and Vleduts-Stokolov 1989). BIOSIS assigns CAS Registry Numbers (RNs) to the chemical names identified in this process. In the STN International online system, hosted in the United States by CAS, a user of BIOSIS can select one or more of the records resulting from a search and extract the RN. The extracted RN can be applied against the CAS Registry File, which contains more than 21 million substances, including organics, inorganics, biosequences, metals, and alloys. The registry file record for the chemical name, including the link to the synonyms for the chemical name and the structure file itself, can then be accessed. With special tools developed by CAS, the structure can be viewed and manipulated. It can be imported into modeling tools that allow the chemist to manipulate the structure and thereby envision new chemicals. Alternatively, the user can start with any database that contains CAS RNs and extract the resulting RNs to perform a search for complementary bibliographic records in the BIOSIS database.
Linking chemical names to structures using RNs on a large scale is neither inexpensive nor easy. There are two approaches to identifying chemical names in text. Some journal articles include the CAS RN for the major chemicals discussed. In this case, an analysis of the text for the terms “RN,” “CAS RN,” and variations preceding numerics can identify RNs that can be used as a link. Alternatively, a program to identify chemical names in text, similar to that developed by BIOSIS, could be devised. Developing the identification program, as well as searching chemical databases, is costly; however, if the digital library has license agreements for chemistry databases, this type of linkage may be possible. In addition, many organizations have small chemical files of their own that may include RNs and other information of particular relevance to the organization’s research. It may be possible to link to these local databases using methods that are more direct.
Linking Personal Names to Biographical Information
A common type of authority file is the personal name authority, which controls variants of personal names. For example, the Library of Congress Name Authority File (LCNAF) is used to control variant personal names for authors, editors, artists, and others. The Union List of Artist Names (ULAN), developed by the Getty Vocabulary Program, is another example. Name authorities serve as tools for catalogers and indexers. They ensure that the proper form of the name, rather than an unapproved variant, is used and bring together all works by or about the person.
A name authority file can also be used to link a bibliographic record or document containing the person’s name to a variety of other related materials. If the digital library’s resource has a standardized form of the name, it can be identified and searched against the authority file to locate variants. The standardized and variant forms can be joined in a search against a variety of other resources that can provide related information.
For example, in the case of a digital library of images of artists’ works or biographical or critical text, a name authority file such as the ULAN or the LCNAF can act as an intermediate file to provide additional information. The file, which contains integrated variant names, can be searched by the name appearing in the digital library collection. When the record is found, the information about the artist can be displayed, providing a wide range of contextual material for the user. Citations to significant biographical or critical works about the artist, some of which may also be available on the Web, may also be provided in the name authority file.
The variant names from a name authority can also be used to locate and provide automatic links from the personal name in the text to a biography, without requiring that the name be presented in the same fashion in the two resources. One such resource that could be linked to for biographical information is Gale’s Biography Resource, which contains more than 142,000 biographies and related citations from more than 1,000 periodicals.
However, to produce this kind of link, there must be a mechanism for locating personal names in text. Several programs can do this type of text analysis; among those that have been developed commercially are NameFinder from the Carnegie Group and the Intelligent Agent from IBM. In addition, variant names can be extracted from the name authority itself, grouped, and run as a search against the text to locate name occurrences.
Linking Entity Names to Physical Specimens
In some cases, it is possible to go another step and connect entity names in the digital library resources to physical specimens. The curation of physical specimens or artifacts is critical to the advancement of many disciplines. Exhibition catalogs describe the art objects in a particular exhibition. Museum catalogs provide inventories of the art, natural history, or cultural objects held by a particular museum. These catalogs, increasingly available as computerized databases, are knowledge organization systems that not only provide descriptive records but also point to the location of the object in a museum, an archive, or another collection.
For example, in biology, a physical specimen is particularly important when it is the result of the discovery and description of a new organism or of the reclassification of a known organism. A type specimen is the example collected from the field by a taxonomist to serve as the prime example for the description of the organism and the validation of its taxonomic classification and naming. These specimens are held by natural history collections, and their deposit is required by the rules of various taxonomic societies.
As part of the curatorial activity, the collections assign identification codes. While the primary use of identification codes has been to organize the physical collections, numerous projects are under way in the natural history community to digitize photographs of specimens and create database records for the specimens, including their identifiers, and thereby make them more readily accessible. The degree of digitization varies from specialty to specialty. For example, in botany, virtually all significant research herbaria are digitally cataloging their type collections instead of maintaining paper records. Many are also making digital photographs of the type specimens available over the Web.
The publication of identification codes in the journal literature is also changing. Historically, identification codes have been presented in the “Materials Used” sections of journal articles. The level of specificity of the identification code has varied, depending on the biological discipline. For example, botanical journals tend to list only the institution and the catalog, while vertebrate journals provide the code to the specimen level. The current trend is to require lists of specimens that are more detailed. As the lists become longer and the printing costs increase, journal publishers are beginning to request links to independent Web sites maintained by the researchers or their organizations that carry all the specimens used in the study and provide some level of identification.
If the digital library collection contains resources that include the identification codes, these codes can be extracted and matched against the Web-based catalogs or databases. This link can provide users with location and contact information to allow them to access the physical object mentioned in the digital library resource.
Curators or registrars of artistic, archaeological, and cultural history collections also assign inventory or accession numbers to items in their collections. Identification numbers may also be found in scholarly catalogues raisonnés. Links similar to those described for natural history can be made between text related to works of art and the physical work in a particular collection. An article about a work of art can be linked to additional information about the physical specimen by linking the identification number in the text with an online catalog containing the number and additional information about the work.
As museums digitize their collections to establish a presence on the Web or to reduce the handling of the physical objects, KOSs that can link the digital library resources to the physical object are being developed. If there is a museum with a collection that complements that of the digital library, it is worthwhile to discuss ways in which the digital library and digital museum collections may “co-evolve.”
Digital libraries can use KOSs to link digital resources to other digital resources or, indirectly, to physical objects. A simple example is the expansion of codes and acronyms. Descriptive records may also be provided either directly from the KOS or indirectly by using the KOS to capture a search key that can be used to access another resource. This concept may be taken a step further by using a KOS, such as a museum or exhibition catalog, to provide information about the location of the physical object.