Resource Library: Multimedia

home *** CD-ROM | disk | FTP | other *** search

/ Resource Library: Multimedia / Resource Library: Multimedia.iso / hypertxt / msdos / hypdiss / chapt2 < prev next >

Wrap

Text File | 1992-04-11 | 78.4 KB | 1,342 lines

CHAPTER II REVIEW OF THE LITERATURE Introductory Comments Since the late 1980s, hypertext investigators have proclaimed this technology as a major advance in electronic information retrieval and publication systems. Yet as of mid-1991, the Compact Disk-Read-Only-Memory (CD-ROM) version of The Oxford English Dictionary on Compact Disk (1987) <refs -oxford> is the most significant of the few commercially distributed hypertext "publications" or products. Hypertext literature and research has reflected a primary orientation toward the technology of implementation. There has been much attention to the subject of optimal user-interface design and to advances in hypertext display technology. There has been little attention in this country to the information organizational and information access concepts used in hypertext system building. The first section of this chapter examines the literature concerning the broader background of established library and information retrieval techniques. This section will not consider the specifics of information retrieval system design and operations. It instead examines the general background and development of information retrieval system concepts and design. The literature included here will be more concerned with "exactly what are we trying to do?" questions, rather than with the specifics of information system implementation. The review will present an overview of major contributors and principles of organizing or structuring access to information. The second section of the chapter focuses on the hypertext literature. It presents a general picture of current views of hypertext information systems. Similarly, it does not cover specific implementation or technical details. Approach to the Literature Search The literature search covered the two distinct areas pertinent to establishing a context for the study: 1) general information access methodology and 2) basic hypertext characteristics. In reviewing the general access topic, the writer searched for broad descriptions or characterizations of traditional information storage, retrieval, and organization approaches. The monographic literature was the chief source for this global view, rather than the fragmented approach found in the journal literature. Accordingly, the search concentrated upon substantive, synthesized works by major contributors and upon topical collections. Review articles and other evaluative citations furnished many candidates. The writer also solicited recommendations from dissertation committee members about major contributors in the area. Hypertext is still discussed primarily in the journal literature. There are at present few substantial monographic treatments. The writer's examination of the published literature confirmed that no major study of general organizational approach or information access design of hypertext systems has yet appeared. The hypertext topical literature search included automated searches of ABI/INFORM, COMPENDEX, Computer Select, ERIC, InfoTrac, INSPEC, Library Literature, and Science Citation Index. In addition, the writer made extensive manual followup of papers and related citations published in the proceedings of Hypertext '87 and Hypertext '89, two major conferences on this relatively new area of interest. Traditional Information Access Approaches Library science literature concentrates upon the objectives and methods of information organization and access design. These questions are after all the central concerns of the discipline. Yet, the culturally and pragmatically oriented views of information access are two very different approaches to considering the role of graphic information storage and retrieval. Reflecting the first approach, Shera (1976) <refs 20 20> points out that a culture needs an effective medium for the transmission of thought. He feels that the various communication media, with an emphasis upon written communication, serve as the binding force that shapes cultures as a cohesive whole. He postulates that communication is the central mechanism linking the cultural components which he identifies as: social organization, tools (technology), and scholarship. He thus stresses the cultural contributions of library and information systems, in providing efficient access to recorded information (Shera, 1976). Many other major writers similarly emphasize the cultural and historic importance of libraries and organized information retrieval systems (Johnson and Harris 1976 <refs -johnson>; Rowley 1987 <refs -rowley>; Foskett 1982 <refs -foskett>). Writers expressing the second view, the pragmatic or practical aspects of information retrieval, do not contradict or diminish the cultural role. Instead, they rather emphasize the functionality of information retrieval. Shera himself concedes that the primary professional responsibility of the librarian and information retrieval specialist is "that of bringing recorded knowledge under manageable control" (Shera 1976, 59) <refs -shera>. There are numerous expressions of the pragmatic view. For example, Kent (1971) <refs -kent> notes that the utopian objective of an information retrieval system is to deliver pertinent information in response to a reasonable question. Koenig (1983) <refs -koenig> and Rowley (1987) <refs -rowley> concur with this goal-centered view. They portray the function of information system design as a pragmatic matter of efficient organization and location of stored information or knowledge. In Koenig's words, ". . . the central thread is knowing how to put information away and find it again" (1983, 193). Borko and Bernier point out that knowledge organizing methods are central and absolutely necessary to the whole operation of our society. They note that we are so accustomed to organizational and information access tools that we habitually and unconsciously depend upon them. The authors refer not only to specialized tools like library catalogs, indexes, and databases, but also to more prosaic information retrieval systems, such as telephone directories, school curriculum numbering systems, product catalogs, purchase order systems, posted building tenant registers, and bus routing tables. They observe that our culture takes such information retrieval mechanisms unconsciously and completely for granted, viewing them as basic infrastructure for living (Borko and Bernier 1978) <refs -borko>. Major Contributors to Traditional Information Access Theory Within the two major schools of thought, several distinct approaches to subject access to knowledge have been advocated. Dewey and Cutter were two early and important innovators of methods of organizing access to recorded information. Dewey is widely known for devising and popularizing his Decimal Classification, which remains an important and widely used bibliographic organizing system to the present day. He himself noted that although he had produced an efficient coding and shelving scheme, it was not at all rigidly based in abstract classification theory. Instead it was actually an arbitrary and subjective standard format for effectively depicting the organization of knowledge. He wrote that he had rejected theory in favor of consistency, mnemonic convenience, and practical usefulness (Dewey 1885) <refs -dewey>. Dewey successfully presented his classification as a foundation and operating standard for pragmatic library practice. He correctly predicted the practical efficiencies and benefits to be gained by adherence of many libraries to a single, compatible, organizational and shelving scheme. These benefits of standardization definitely contributed to the widespread adoption of the Decimal Classification. Dewey also provided additional advantages by incorporating his classification methodology into his formal library school offerings. Last but not least, his considerable personal skills in marketing and promoting the classification were important factors in the system's success (Sellberg 1988 <refs -sellberg>, Shera 1966 <refs -shera>). Cutter's 1904 publication Rules for a Dictionary Catalog is still considered one of the major expressions of the subject classification approach (Shera 1966) <refs -shera>. His major contributions, such as uniformity in headings, preference for common usage, portrayal of knowledge as reflected in the literature, and the principle of specific and direct entry, remain important concepts of classification to the present time. Cutter's definition of classification was demonstrably the major influence upon the design of the Library of Congress (LC) Classification (Chan, Richmond, and Svenonius 1985) <refs 4 20>. As Dewey suggested, the attractions of standardization played a large part in the adoption of the LC classification. The LC system also became a major classification standard, due in large part to the cooperative cataloging economies, and to the attractiveness of centralized catalog card publishing and distribution. Cutter's approach to subject classification is also evident in both the Sears subject headings and the Wilson index publications (Foskett 1982 <refs -foskett>; Rowley 1987 <refs -rowley>). In his work on Systematic Indexing, Kaiser contributed the first consistent approach to the order of presentation of the aspects of composite subjects (Rowley 1987) <refs -rowley>. He proposed that the relationships of composite subjects should be consistently expressed in the order of first Concrete, then Process. Kaiser gave specific recommendations for the treatment of complex topics, and for handling geographic subdivision or place names. His simple concrete-process approach was the basis for many modern subject representation systems. Kaiser's system has been successively extended by Ranganathan, Coates, and Vickery, for the handling of more complex subject representations (Foskett 1982) <refs -foskett>. There have been other enumerative classification schemes or theories of good quality and theoretical base. Those of Bliss and Sayers are good examples (Foskett 1982) <refs -foskett>. However, it appears that the acceleration of growth and size of recorded knowledge brought an end to the development and adoption of the great enumerative classification schemes. A main drawback of these massive organizational schemes was that they were essentially static. Enumerative schemes are simply not able to satisfactorily and efficiently reflect the constant shift and expansion of modern knowledge (Foskett 1982) <refs -foskett>. Enumerative systems contain a basic theoretical difficulty: the impossibility of defining consistent principles of subdivision in a universal, hierarchical classification. In Britain, after extensive deliberation, the Classification Research Group concluded that schemes using the enumerative approach were inevitably doomed to failure, due to inconsistency and high maintenance expense. They finally decided that it was impractical to keep such a representation current with the development of modern knowledge (Classification Research Group 1985) <refs -group>. Ranganathan's facet analysis concept, expressed in the 1930's, was the next major advance in the approach to subject classification. Facet analysis is defined as a controlled process by which a document is distilled into its component subject aspects or categories. These aspects or facets can then be used to synthesize a classification code by stringing together or connecting the class identifiers of the basic concepts. Ranganathan's system depends heavily upon punctuation to express the complex relationships between the various facet aspects. This is in contrast to enumerative approaches like Dewey and Bliss, which are mostly prestructured prior to use (Cleveland and Cleveland 1990) <refs -cleveland>. Ranganathan's scheme was in itself an evolutionary synthesis, since it carried on or developed concepts expressed or suggested by Gesner, Cutter, and even to some extent by Dewey (Shera 1966) <refs -shera>. The aspect analysis central to Ranganathan's facet approach is particularly efficient when combined with mechanized manipulation and synthesis. He developed his subject classification methods within the traditional, manual processing environment. The controlled concept vocabulary and the standardized methods for aspect and relationship representation were intended to produce synthesized classification coding strings. These synthesized strings were then to be used for production of printed catalog listings or indexes, and for document collection arrangement. However, the "decomposition" effect of facet analysis is also a pragmatic and economical approach for description or representation of document content, using a controlled collection of concept terms (Chan, Richmond, and Svenonius 1985) <refs 4 20>. Shera points out that the aspect analysis approach has so far proved to be the most efficient for mechanized information retrieval. He observes that limitations in the creation and use of the overly complex synthesized facet codings have been evaded or sidestepped by the advantages of machine search and manipulation efficiencies (Shera 1966) <refs -shera>. Ranganathan's facet identification approach was also complimentary to the ideas of one of the pioneers of the computerized retrieval approach. Taube introduced the use of "descriptor" indexing, which is similar to the facet concept of subject component representation. But instead of using the subject descriptors to synthesize classification expressions, Taube specifically intended them to be used for computer searching and manipulation. He saw and championed the potential of Boolean logic in postcoordinate searching, and of computer processing and manipulation for bibliographic processing (Taube 1985) <refs -taube>. Taube's early proposals centered on the use of simple "Uniterm" descriptors. These were single terms for expressing subject content which were to be directly extracted from source document text. He originally advocated the exclusive use of simple term extraction to build vocabularies of single terms. He did not include the use of vocabulary control methods. However, Taube soon found it necessary to adopt vocabulary control methods, like rules for syntax and control of synonyms. These measures included pre-coordinated or multiple-word terms, to better express complex concepts, and to give indication of relationships (Foskett 1982) <refs -foskett>. Taube made early demonstration of the potential power of computerized retrieval. His experience showed the advisability of combining the effectiveness of controlled vocabulary with the efficiency and synthesizing power of Boolean post-coordinate retrieval. His innovations led directly to the current mainstream approaches employing bibliographic and text information retrieval. He also clearly identified the synergistic effects to be gained from machine manipulation of bibliographic information: The system of bibliographic coordination has three major advantages over the systems of classifying and indexing. In the first place, a system of bibliographic coordination is designed for machine sorting and collating. . . . Secondly, once the categories have been determined, the actual indexing of material in any particular field would be very simple. . . . Finally, the system of bibliographic coordination would, if properly set up, disclose to the searcher more information than had been put into the system. Similarly, in the system of bibliographic coordination, by combining various categories, we may get information which, although implicit in the system, was never explicitly recognized. (Taube 1985, 123) <refs -taube> Text string manipulation has been another major approach in modern subject access techniques, since the advent of automated processing. String manipulation uses automated processing of titles, descriptions, and subject codings, to produce multiple versions or permutations of the subject term or aspect identifiers. The primary leverage of string manipulation in subject access application is gained by taking advantage of the power and economy of computer processing. Luhn directed the first major implementation using string manipulation of document titles. His system produced permuted index listings of titles in KeyWord In Context (KWIC) format (Luhn 1985) <refs -luhn>. His KWIC indexes were printed alphabetical listings, with substantive or sorting words listed down the center of the printed index lines, for easy recognition. The context of the word's occurrence was printed on either side of the word, in order to communicate the document title context. Titles were "wrapped" or rotated, as needed, so the context would fill the line, with the subject word remaining in the center position. Luhn admitted this to be an imperfect approach to subject identification or representation. It ignored such problems as vocabulary consistency, synonymy, and homonymy. It depended completely upon the author's skill and desire to effectively express his or her topic in the document title. But he also pointed out that KWIC indexing eliminated the substantial investment necessary for traditional manual indexing operations. Luhn advocated this indexing method as a fast, cost- effective, approach for communication of the content of technical literature. He in fact portayed it as leaning substantially towards the goals of technical literature communication or distribution, rather than functioning primarily as an information retrieval tool (Luhn 1985) <refs -luhn>. Responding to information science's penchant for devising systems for "automatic subject content analysis," Cleverdon (1985) <refs -cleverdon> has stated that conceptual indexing can not be entirely avoided. In this respect, he points out, KWIC indexing does not really avoid the necessity for conceptual indexing, it merely delegates the task to the original author of the document. As a final expression of his facet theory, Ranganathan devised the Colon Classification, a complex methodology for synthesis, or construction, of subject classifications. Separate facet codes were concatenated, using strict and complicated sequencing and punctuation rules. These constructed codes could accurately portray the complex relationships of the component aspects. However, in both construction and interpretation, the complex, dense, code strings were quite difficult to comprehend. It later became evident that the complex sequencing and construction logic could be automated. System designers could thus make use of computer assistance to make human classifier operations easier and more efficient. To this end, the Classification Research Group (CRG) undertook detailed consideration of various approaches for document analysis and access during the 1950s and 1960s. They correctly anticipated wide application of automation for indexing purposes. As noted, they decided to incorporate the facet analysis approach at an early point in their investigations (Foskett 1982) <refs -foskett>. In the late 1960s, Austin directed the creation of what became the PRECIS classification system, using the CRG's preferred facet approach. From the start, PRECIS was a highly practical project, since it was intended to be used for production of the printed British National Bibliography. The PRECIS system successfully incorporated facet analysis and controlled vocabulary approaches. It included representation of relational, hierarchical, and geographic identification, for accurate communication of complex concept relationships. The final computerized implementations efficiently produced the constructed strings. The basic strings were then automatically permuted or "rotated" into multiple likely forms, for filing in printed index listings (Austin and Digger, 1977) <refs -austin>. Citation indexing is the contemporary subject access approach bearing perhaps the strongest relationship to the hypertext approach. Citation indexes are lists of documents, with a sublist under each item of subsequently published papers that cite or refer to the indexed item (Garfield 1985) <refs -garfield>. As Cleveland and Cleveland have observed, "The premise of citation indexing is that citations reflect document content, and that an author's citations can therefore be substituted for the judgment of indexers" (1990, 72) <refs -cleveland>. The citation index is therefore a perfect reflection of the internal structure of the domain literature. Although other factors should be considered in evaluating this index type, the citation index has been widely accepted and adopted by end users. Many of them prefer citation indexes over more traditional bibliographic research tools. In more traditional index approaches, a researcher must first make the unfamiliar and demanding mental effort of exactly defining his or her informational needs. These must then be translated into the format vocabulary and syntax of the particular information access system. The need for this whole process may be viewed as a substantial obstacle to index use, since end users are unused and unskilled in expressing subject or information need concepts in formal access vocabulary terms. On the contrary, citation indexes are easily approachable by end users. They can use the obvious and easy entry point of "a document or author of known interest." Citation indexes are accordingly quite popular with end users within many disciplines. This user preference is similar to the oft-expressed preference for hypertext use (Nielsen 1990) <refs 18 20>. This natural human preference for the course of least resistance has been expressed by Zipf (1972) <refs -zipf> as the Principle of Least Effort. Björklund (1990a) <refs 2 21> similarly cites Krikelas as noting the tendency of information seekers to specifically base their search activities upon some concept of minimum personal effort. Printed citation indexes would appear to be similar to the hypertext associative approach. Much like hypertext, citation indexes capitalize upon user recognition of topical relevance, using context or example. Colloquially phrased, both approaches capitalize upon the user's "I knows it when I sees it" recognition of relevance. They also both avoid the mental strain of the user needing to learn how to deal with the information retrieval system. In the same way that footnotes within a document contain the author's judgment of what documents are relevant to his or her topic, the embedded links within an authored hypertext document contain the same sort of relevance judgment. The links are clearly visible, in this case, as a suggestion for the nearly effortless hypertext associative retrieval commands. Subject Access Overview Works Broad overviews of the subject access topic appear in both works giving an introduction or basic survey, as well as in works taking reflective or philosophical views. A systems analysis or engineering orientation has also emerged in recent years. Borko and Bernier (1978) <refs -borko>, Rowley (1987) <refs -rowley>, and Cleveland and Cleveland (1990) <refs -cleveland> are basic introductions to indexing and abstracting. All three books offer an historical view of subject indexing development, of the various approaches to and types of indexing, and of practical indexing operations. Metcalfe (1957) <refs -metcalfe> offers a theoretical, reflective, and highly personal approach to subject access practice. He presents detailed historical review of the evolution of subject representation in indexing and classification. There is extensive analysis of minute developments and changes in the approaches of the major figures. Writing in 1957, he covers only the earliest ventures into "mechanical retrieval," including Taube's "Uniterm," Mooer's "Zatocoding," and Shaw's "Rapid Selector." But the detailed analysis of these early systems can certainly be of interest to modern readers. Metcalfe examines the metaphysics, logic, and philosophical approaches of the major cataloguing and classification figures. He is highly critical of classified approaches to cataloging and indexing, including Ranganathan's analytico-synthetic approach (Foskett 1982) <refs -foskett>. He blames much of the weakness of British bibliographic tools, compared to American products, to the British adherence to classified approaches. Metcalfe's book is a highly readable and informative one, useful to those wishing to examine the origins and evolution of modern subject access approaches. Foskett (1982) <refs -foskett> presents a balanced overview of subject access theory and historical development, comparable in scope to Metcalfe's coverage. He operates from the basic social or cultural view of the library role, as the information transfer agency. He distinguishes knowledge and information as follows: "knowledge is what I know[,] information is what we know." (1982, 1) He then expresses the view of libraries as a fundamental link in the chain of communication of knowledge: . . . society began to move forward when information of various kinds began to be recorded in relatively permanent forms which could serve as substitute for the wise man in person. Knowledge only becomes generally useful when it is made available; by recording it, we do our best to ensure that it is permanently available to anyone who may need it, instead of ephemeral and limited to one individual. (1) Foskett bases his presentation upon this need for effective communication of information. His book is important in its coverage of the historical and theoretical bases for subject access systems. It deals with the development of information retrieval automation, methods of information systems design, and quantitative systems of evaluation. His treatment, however, emphasizes the goal of human communication, not the purely mechanistic or technical view of systems design. Foskett uses a wide combination of approaches to give a comprehensive and balanced view of the topic. Several other writers concentrate more on the systems analysis style of approach. For instance, Meadow (1973) <refs -meadow> gives a clearly stated and logical view of information retrieval systems. His treatment arises very much from the systems viewpoint and from the Shannon and Weaver communications model (Shannon and Weaver 1959) <refs -shannon>. This early communications model reflected a technical, "effective transmission" view, rather than a social communication function. Meadow summarizes the problems of indexing and subject representation and describes the necessity for formalized index language solutions. His is a capable, but rather task-oriented treatment. As he himself straightforwardly phrases it, ". . . information retrieval means the retrieval of information-bearing symbols from a file, and is independent of the nature or form of the information retrieved or the use to be made of it" (1973, 120). Vickery (1987) <refs -vickery> presents a view of information sciences as related to communication in society. It is also more of a descriptive, objective view, without the "wisdom-centeredness" characterized in Foskett's presentation. Vickery is firmly within the tradition of alphabetical indexing theory, but he has not chosen to emphasize subject access theory in this work. It is rather a broad treatment of the entire area of information science. He does, however, briefly note the research or theoretical traditions of subject access, characterizing them into four main themes. He describes these as: the structure of classifications; the tradition of alphabetical indexing; the interpretation of classifications and indexes as retrieval languages; and the newer approaches arising from the computerization of information. Another group of writers expresses a highly pragmatic or "engineering problem" approach. Milstead (1984) <refs -milstead> capably presents the historical basis for subject access system approaches, but she emphasizes the system design aspects of the question. She depicts the system design problem as a set of interacting choices, describing the ways in which one choice may constrain others. (1984, 5) She sympathizes with the human purposes for knowledge access, but she defines primary goals very objectively, as follows: 1) minimize total time and effort to find answers; 2) minimize redundant content; and 3) minimize the number of lookups. (17) Milstead effectively presents her problem-centered viewpoint. She describes the design of a subject access system not only as a subjective or theoretical exercise, but also as a practical exercise in cost-effectiveness. In her view, the efficacy of a final design is highly dependent upon the wise selection of trade-offs. In another pragmatic viewpoint of information system planning, Taylor (1986) <refs -taylor> presents his "value-added model" of information systems. This approach analyzes the values an information system delivers to users. The model does not address the specific details of subject access approaches. It rather attempts to define the values that users find to be important in selecting information systems. Taylor generalizes the value-adding activities of the systems as: 1) selecting input; 2) storing and organizing for physical access; 3) describing and labeling for retrieval; 4) analyzing, evaluating, and comparing for quality; 5) combining and interpreting; 6) responding and adapting to specific user problems. (1986, 202) The results of many of the activities will of course be dependent upon the success of the document representation and delivery mechanisms. In Taylor's phrasing of the required operations, the system: * displays its contents, either in menu/instruction form or in text as direct response to a query; * transmits signals concerning the potential value of portions of its contents; * provides a variety of means to aid the user in making choices. (1986, 54) Although these are very generally stated goals, they are obviously specifically applicable to the design of a good subject access system. Taylor's treatment is thought- provoking, and may be applied to a computerized library catalog, an online fulltext database, or a hypertext system. It is a well-reasoned approach to answering the aforementioned "what are we trying to do here?" question. Hypertext System Background This section will identify important early implementations using hypertext information system approaches. It will also cover general hypertext background. The review will not attempt to cover technical details, nor details of specific implementations, but will instead concentrate upon hypertext system concepts pertinent to this paper. The section will therefore not reflect the common literature emphasis upon hypertext technology. Historical Development As mentioned in the introduction, Bush initially proposed the hypertext concept in 1945. However, except for his imaginative vision, there was no substantial progress until the 1960s. Engelbart's NLS/Augment project began at Stanford in 1962, including components of office automation, text processing, and hypertext, and was publicly demonstrated in 1968 (Engelbart, 1988) <refs -engelbart>. The Augment system, later marketed by McDonnell-Douglas, has supported a group of over 1000 knowledge workers for over twenty years (Horn 1989) <refs -horn>. Nelson, coiner of the term "hypertext," was an early hypertext visionary and promoter of an ambitious global network hypertext system concept called Xanadu. This is conceived as a distributed network of back end local hypertext databases (Nelson 1974) <refs 18 17>. Nelson proposes a repository system to hold everything ever written, and being written, with intricate hypertext connections between all related documents. Xanadu is now supported by the substantial resources of the Autodesk software company. The Xanadu product has not been released at the time of this writing, although Nelson predicts it will be made available during 1991 (Ditlea, 1990) <refs -ditlea>. Andries van Dam directed the next major hypertext system activity, beginning in 1967 at Brown University. The Hypertext Editing System was the world's first operational hypertext system, and was eventually installed at the Houston Manned Spacecraft Center. In 1968, van Dam then implemented the File Retrieval and Editing System (FRESS) hypertext system at Brown University (Nielsen, 1990) <refs 18 20>. Brown University remains a major center in hypertext system development, with its Intermedia hypertext implementation. This system is a highly integrated system, with editing, calculating, text searching, dictionary consultation, graphic/animation, and hypertext abilities. It is used as a general teaching and learning environment, in curriculum support, for individual communication and as a channel for collaboration (Yankelovitch, 1985) <refs 25 7>. The University of Maryland has also been active in hypertext system research and development. Shneiderman implemented the TIES (The Interactive Encyclopedia System) project at this university in the early 1980s. The system has since been renamed to Hyperties (Shneiderman, 1987b) <refs 21 4>. Shneiderman is active in the research areas of software psychology and user interface research. As a result, there has been much investigation in these particular areas of the hypertext system investigations at the University of Maryland. Microelectronics and Computer Technology Corporation (MCC), the American computer technology research consortium, has been another major hypertext research center. Conklin and Begeman have reported upon the "gIBIS" system developed at MCC. (The system acronym is derived from "graphic, issue-based information system.") gIBIS is a networked hypertext system designed primarily for the coordination, tracking, and facilitation of software development projects. The hypertext network was designed as a system to identify issues, enhance communication, and track issue resolution, for large software development teams with software personnel that may number in the hundreds (Conklin and Begeman 1989) <refs 6 4>. Hypertext Overviews and Bibliographies Conklin (1987) <refs 6 1> published the earliest comprehensive and definitive piece on hypertext. He described system concepts, types and subtypes of hypertext link mechanisms, characterized the variety of system approaches, and surveyed existing implementations. His type description may be paraphrased as: * Macro literary systems, large online libraries, of the Xanadu model; * Problem exploration tools, for personal and group problem exploration, outlining, authoring, and design; * General hypertext technology, general purpose hypertext implementations, for experimentation and innovation; * Structured browsing systems, designed for mass user information systems, information retrieval and dissemination, and electronic publication. This paper is most concerned with the final category mentioned above. "Structured browsing systems" will be the basis for "publication" systems, for mass market and vertical market information distribution. As such, they must incorporate highly effective information retrieval methods. The systems will need to simultaneously provide both power user sophistication and simplicity of use for absolute novices. Conklin also described the "lost in space" phenomenon, the hypertext system problem most commonly mentioned by writers. This confusion of orientation or location is caused by the chaotic maze of linked network nodes often presented to the user in a poorly designed hypertext system. This creates "potential for the user to become lost or disoriented" (1987, 38) <refs 6 1>. As a computer professional, Conklin perhaps predictably suggested technological solutions. First, he recommended "graphic browsers," or pictorial or diagrammatic representations of local or global link connections; second, he suggested improvements in keyword or text search. Fiderio (1988) <refs -fiderio> offered another comprehensive introduction to hypertext, more generally directed, and much less technical than the Conklin overview. Shneiderman and Kearsley (1988) <refs 21 16>, and Jonassen (1989) <refs 15 2> also contributed popularized book-length introductions to the hypertext topic. Barrett, Berk and Devlin, Horn, and Martin have produced works concentrating upon the "editorial management" of hypertext, covering implementation or production of hypertext information systems. Barrett's book collected contributions from active investigators and developers of hypertext systems, describing practical experience, and identifying problems in production and design (Barrett 1989) <refs 2 7>. Horn and Martin presented detailed guides to structured, efficient, hypertext design and production management (Horn 1989 <refs -horn>, Martin 1990) <refs 17 12>. Berk and Devlin compiled a substantial reference guide to the design and implementation of hypertext systems, with contributions by many highly regarded hypertext developers and authors. Their work includes chapters on editorial design and authoring tools, production details, legal issues, and case studies (Berk and Devlin 1991) <refs -berk>. Hypertext '87 and Hypertext '89 are proceedings of two major conferences on the topic, sponsored by the Association for Computing Machinery (ACM). They are major collections of papers on the subject, with multidisciplinary approaches (Association for Computing Machinery 1987 <refs 1 11>, 1989 <refs 1 16>). These proceedings were substantial early collections of papers, serving to centralize work on the topic in the absence of monographs during this period. Trigg and Lin have produced dissertations on hypertext. Trigg's early work (1983) <refs -trigg> was probably the first dissertation on the topic of hypertext. He described the TEXTNET system, used for online scientifc document and information sharing. Lin (1989) <refs 16 18> addressed the question of providing multiple metaphors for hypertext information access. As in most other hypertext research, the term "metaphor" was here applied to the general concept of presentation interface. Lin's investigation utilized a tourist information system, with presentation of information via book, note card, and map interface metaphors. In contrast to Lin's study, this paper does not focus on presentation metaphors. It is concerned rather with hypertext incorporation of formal or traditional information access presentation methods, such as indexing, hierarchical classification, or string-searching approaches. Nielsen (1990) <refs 18 20> offers the most comprehensive book treatment of the topic. In it, he offers detailed coverage of historical background, implementation descriptions, applications, system architecture, usability, evaluations, and predictions for future development of hypertext. In his book, Nielsen provides a selective annotated bibliography, of high quality. Knee and Atkinson (1990) <refs -knee> have also produced a comprehensive, book-length, annotated bibliography on the hypertext system topic. The two works together offer substantial bibliographic resources on the topic. Educational Applications of Hypertext There has been considerable literature on the educational applications of hypertext system. These have ranged from lower schools to university levels (Beck and Spicer 1988 <refs -spicer>; Chignell and Lacy 1988 <refs -chignell>; Crane and Mylonas 1988 <refs -crane,>, Smith 1988 <refs 21 21>; Underwood 1988 <refs -underwood>; Jonassen 1986 <refs 14 22>, 1989 <refs 15 2>; Marchionini 1988 <refs -marchion>; Yankelovitch, Landow and Cody 1987 <refs 25 3>). Many of these writers stress the advantages of the connectivity of knowledge and the effectiveness of the learner-driven presentation of information. Locatis (1987) <refs -locatis> notes the attractiveness of this self- directed educational approach: Each display in the knowledge base functions as a 'menu' to other displays, and contains cues to additional information available. . . . [Learners can] mark displays for later retrieval, for referencing personal notes, or for evoking searches. (1987, 94) He summarizes the advantages and integrative aspects of hypertext for educational applications. His list includes: storage capacity, integrated media capability, interlinking of information systems, and fast information retrieval. Most cited writers do not consider information organization or structure in depth, but instead focus upon specific hypertext authoring styles. They usually recommend a structured or hierarchical approach, but there is generally no depth of treatment of how to best achieve this. Instead, most writers appear captivated by the flexibility of hypertext linking and the resulting fluidity of style afforded to the author. The apparent subjective lure of hypertext systems has affected educational implementations at all levels. The Brown University Intermedia system (Yankelovitch et al. 1988) <refs 24 22>, and the University of Southern California's Project Jefferson (Chignell and Lacy 1988 <refs -chignell>; Kinnell 1989 <refs -kinnell>) are advanced investigations and implementations of hypertext technology for academic and research infrastructure purposes. The various reports on these two projects discuss the interdepartmental and library cooperation in designing the system, and the emphasis upon integration into the teaching curriculum. In both applications, the educational and writing tool priorities appear to have overshadowed the information access elements. Importance of the User Interface The "user interface" is a basic concern of computer and information system designers, with much justification. The user interface is the meeting place of users and computer systems; "where the rubber meets the road," so to speak. Flaws or mistakes in human factors design or in effective communication of system messages will have substantial negative effects upon system usability and performance. Computer professionals realize this. They consider the computer-human interface so important that managers now routinely devote as much as one-half or more of total software project cost to user interface design efforts (Baecker and Buxton 1987) <refs -baecker>. Study and design of the "computer-human interface" is presently an active interest area. Reflecting this, the Special Interest Group on the Computer-Human Interface (SIGCHI) is a sizable and active group within the Association for Computing Machinery. The multi- disciplinary concerns have drawn much attention and participation from a number of peripheral specialties. Much literature directly addresses computer-human interface issues. Baecker and Buxton (1987) <refs -baecker> have produced a substantial introductory collection illustrating the breadth of multi-disciplinary concerns relating to human interface design. They have compiled a wide selection of primary source texts, arranged in topical chapters, which results in an excellent introduction to the topic. Helander (1988) <refs -helander> edited a comprehensive handbook on the topic of human-computer interaction, with contributions from a number of approaches from a number of disciplines. Shneiderman and Brown have published highly-regarded standard works on design for the human interface (Shneiderman 1987a <refs 20 24>; Brown 1988 <refs 4 7>). Galitz has produced a handbook specifically covering display screen design. He reviews research showing that interface design features can make as much as 20% difference in human processing speed, interpretation, and errors. This will obviously have significant effects on long-term costs relating to usage, accuracy, and labor (Galitz 1989) <refs -galitz>. Dumas (1988) <refs -dumas> also thoroughly covers user interface design. He presents detailed information on transaction control, data entry, screen presentation, and other topics. Dumas advises on management of the interface design project. Ravden and Johnson (1989) <refs -ravden> offer a practical guidebook to evaluation of the human-computer interface, stressing usability and accuracy. Their guidebook gives a specific checklist approach, along with detailed recommendations for planning, execution, and analysis of the evaluation. As well as clear textual communication and good conceptual metaphors, professionalism in graphic design is critical to interface effectiveness. Faiola emphasizes the need for involving skilled graphic technologists in computer system design, and includes a comprehensive visual design bibliography (1988) <refs -faiola>. Referring specifically to hypermedia, Vainio-Larsson (1986) <refs -vainio> expresses the importance of the critical human interface point. His comments may be extended to software in general. He points out that the user must be able to quickly establish conceptual understanding of the system behavior and actions, in order to attain effective interaction. He describes system/human communication is a "top-down hierarchy," with the user at the top of the hierarchy, communicating via the user interface, to the application. Vainio-Larsson describes a good interface as functioning to shield the user from "implementation close" interactions. This means the user is protected from the necessity of communicating in computer-type language. Hutchins, Holland, and Norman (1986) <refs -hutchins> similarly portray the interface as the user's tool for dealing with the immense distance, or "gulf of execution," between the user, the system internal operations, and the system goals. Library and information science professionals are quite aware of the importance of the user interface. There have been several literature reviews within the library and information science literature, dealing with psychological and human factors interface research (Borgman 1984 <refs -borgman>; Ramsey and Grimes 1986 <refs -ramsey>). Hypertext Performance Researchers are demonstrating that hypertext systems can offer information retrieval and task performance advantages. As suggested in the preceding section, researchers are also finding that interface design significantly impacts upon hypertext system performance. Egan and his fellow Bellcore researchers have shown in a number of investigations that subjects using their SuperBook hypertext system as an information retrieval and research aid achieved substantially higher performance than subjects using equivalent printed sources. They concluded, "Students using SuperBook answered more search questions correctly, wrote higher quality 'open book' essays, and recalled certain incidental information better than students using the conventional text" (Egan et al. Behavioral Evaluation. 1989, 205) <refs 8 1>. The Bellcore team used the "iterative implementation" system design approach. Each evaluation of the hypertext system was followed by incremental improvements, each level being treated as a prototype for the succeeding version. Improvements in the three versions of the system, directed mainly to streamlining the interface, produced substantial performance gains at each level. The final, optimized, "MiteyBook" version especially demonstrated the performance gains from interface improvement. This version delivered 25% better performance than the printed book version, and was significantly faster than SuperBook II (Egan et al. Formative Design Evaluation. 1989 <refs 7 20>; Egan et al. Behavioral Evaluation. 1989 <refs 8 1>; Egan et al. 1991 <refs 8 9>). The United States Army Research Institute conducted an evaluation of the Personal Electronic Aid for Maintenance (PEAM) system in 1987 (Wisher and Kincaid 1988) <refs -wisher>. The research team evaluated a hypertext-based electronic delivery system used for technical manual data. The study involved observation of fourteen Army technicians and fifty-six Navy technicians, performing troubleshooting and repair for a total of 610 hours. Army and Navy technicians worked on M1 tank turrets and Sparrow missles, respectively. PEAM users showed an 11:1 advantage in troubleshooting error rate, and a 2:1 advantage in repair error rate. The Navy technicians using PEAM had a 25% time reduction over test subjects using equivalent printed technical manuals. The Army technicians using the PEAM system performed approximately 14% slower than the group with printed manuals. The difference was traceable specifically to an implementation version differing from the Navy version only in that it had longer graphic display times (greater than or equal to fifteen seconds for image retrieval). This research demonstrates the surprising cumulative performance effect resulting from a single, seemingly trivial, interface issue. Choice of the effective presentation "metaphor" is also an important interface issue. Two major studies have addressed this specific question in relation to hypertext systems. Lin (1989) <refs 16 18> studied multiple navigation metaphors within a hypertext implementation. She found that subjects using a single metaphor were most effective in tasks that were suited to the type of exploration afforded by the particular metaphor model. She also found that subjects using multiple metaphors had longer learning curves in achieving effective performance, due to the necessity of learning more navigation approaches. However, as experience increased, users with multiple navigational metaphors available to them began to show a performance advantage. Hammond and Allinson (1988a <refs 11 18>, 1988b <refs 11 23>, 1989 <refs 12 5>) also studied use of multiple hypertext navigation metaphors. Their study used a small hypertext system with thirty-nine screens containing information about the history of York, England. Available metaphors included base hypertext links only (H), a Map facility (M), an Index facility (I), and a guided Tour facility (T). They used five study groups, one using base hypertext only (H); three using base hypertext and one other facility (HM, HI, and HT); and the last using all hypertext navigation facilities (HMIT). The investigators found that all groups performed similarly on testing, with no significant differences. But a surprising result emerged: the groups saw different numbers of screens. There were mean scores of 18.1, 20.3, 18.8, 19.9, and 17.7 screens viewed, by H, HM, HI, HT, and HMIT respectively. Since there was no significant difference in performance, the subjects seeing the least information screens tended to extract more information from them. They thus performed or scored more efficiently than those who saw more screens (1989, 200) <refs 12 5>. The investigators generalized that the subjects with more hypertext navigation tools available for use tended to perform most efficiently in information retrieval tasks. Authoring and Organizational Issues The problems of the simplistic, intuitive, approach to hypertext system building are obvious. Illustrative of the problem, Nielsen summarized the results of a field study observing users of a "small" Guide hypertext document "which could be read in one hour." He reports that even with this small hyperdocument, more than half of the users agreed with the statement, ". . . I was often confused about where I was" (Nielsen 1990, p. 127) <refs 18 20>. In his keynote address for the ACM "Hypertext '87" conference, van Dam covered nine key areas "we all ought to be looking at." Numbers five and seven respectively were the need for improvements in "hyperspace" navigation and the need for better organizing, indexing, and accessing the hypertext knowledge bases (van Dam 1988, 894-895) <refs -van dam>. In a chapter on the topic of authoring hypertext, Shneiderman cautions that "there is a real danger that [authoring] can also lead to hyperchaos. He points out dual dangers: 1) that hypertext may be inappropriate for some projects, and 2) that the design of the hypertext may be poor (e.g., too many links, confusing structure)" (Shneiderman 1989, 116) <refs 21 10>. He advises control measures, which may be viewed as versions of standard library or indexing operation control methods. His suggestions included: starting with a meaningful overall structure concept; consistent naming of documents/nodes; maintaining a master reference/control list; and verifying the existence of nodes named in link references, to avoid blind references (1989, 125) <refs 21 10>. Reporting on the three teams that translated the Hypertext '87 conference proceedings into their respective hypertext products, Alschuler comments on the inconsistency of implementation, the poor indexing, the disorganization of index lists, and the production team reports of extreme difficulties. Elise Yoder, working with the "KMS" hypertext product, told her that it was simply not possible to connect everything as thoroughly and effectively as they had intended. "By the end of the project, in [Yoder's] words, they were 'practically fabricating' meaningful connections in order to install more links" (Alshuler 1989, 358) <refs 1 5>. Shneiderman, of the Hyperties version authoring team, reported to her that the task "confirmed his suspicion that this type of 'linear' text could not be translated into hypertext" (358). These individuals were all experienced hypertext system implementers, working on a relative small knowledge base (The full printed conference proceedings occupy only 432 pages, plus index). Alschuler muses: Was this poor showing due to sloppy, hurried construction rather than the fundamental limitations of the programs? Each group had a team of two or three people and reported spending approximately two months preparing the hypertext publication. . . . [I]t would be impossible to find . . . more qualified individuals than those responsible for the applications. . . . It does not seem likely that lack of development time or talent was an issue. (1989, 358) Alschuler's concern over this apparent paradox is understandable. The writer has prepared a subjective response to this question, which appears in Chapter V of this study. The Engineering Approach In contrast to the intuitive or subjective approaches, there are reports of standardized, methodical, and generally successful approaches to hypertext system creation or conversion. Most of these have been for commercial or institutional publishing purposes, and were based on a task-oriented, engineering outlook. Raymond and Tompa (1988) <refs 19 21> have reported on the conversion task and analysis needed for the monumental translation of the Oxford English Dictionary into a CD-ROM hypertext format. Frisse (1988b) <refs 9 17> covered the planning and task execution for the hypertext conversion of the Washington University Manual of Medical Therapeutics, a 500 page institutional guide for internal medicine standard practice. Glushko (1989 <refs 10 21>, 1990 <refs 11 2>) described the conversion of the 3000-page Engineering Data Compendium into a hypertext CD-ROM. There are also several book-length guides to this more structured or disciplined hypertext system-building approach. Information systems consultant James Martin has published Hyperdocuments & How to Create Them, a detailed guide for developing effective electronic hypertext document systems (1990) <refs 17 12>. He proposes a basically hierarchical structure planning approach, stressing the creation of clear, consistent trails. As does Frisse (1988a) <refs -frisse>, Martin proposes the application of outlining software tools as an aid to definition and structual maintenance of consistent hierarchical hypertext structures. Martin deals extensively with authoring project management, covering printed text conversion, authoring team management, incorporation of graphics, methods of ongoing database management, and other important production topics. Similarly, Horn (1989) <refs -horn> has adapted "Information Mapping," his structured technical writing methodology into a hypertext authoring approach. His approach is also based on highly structured hierarchical design. Similar to Martin's methodical approach, Horn thoroughly covers the spectrum of hypertext information system design and management. His presentation focuses on the authoring and stylistic tasks, but does not cover the general production topics in detail, as does Martin. Neil Larson, the developer of the MaxThink hypertext authoring system, the subject of this case study, also stresses the hierarchical authoring approach (1987a <refs 15 24>, 1987b <refs 16 4>, 1988 <refs 16 7>, 1989 <refs 16 10>, 1990 <refs 16 14>). In fact, his company's first software product was "MaxThink," an award-winning outliner program, which later became one of the cornerstones of the hypertext authoring system. He subsequently produced "Houdini," a more advanced matrix (three-dimensional) outliner program. The matrix outliner is capable of building highly complex network outlines, in which any topic or node can be connected to any other node. The application to effective hypertext system building is obvious. For example, Larson's associate Tony Phillips currently uses the MaxThink system in authoring the DaTa hypertext. This is a CD-ROM hypertext product covering the accounting/auditing subject area. The DaTa hypertext currently (March, 1991) occupies one hundred megabytes of disk space, containing approximately 60,000 screens and more than 150,000 hypertext links. The system is maintained in about 200 highly interconnected hierarchical networks. The complex structure is maintained and updated using the Houdini matrix outliner. The James Martin Report on text management praises the MaxThink structural approach in his publication on text management (Dewire and Locke 1990) <refs -dewire>. They note: It is the ability to categorize and classify ideas in ways that match the needs of users that creates efficient hypertext applications. Hierarchies are a universally understood method of communicating information. A hierarchy is a subset of a network, graph, lattice, or outline. However, in searching for the best hierarchical format for information, a network is often the best place to start. Networks allow users to rapidly classify information by multiple dimensions without forcing them to immediately choose which dimension belongs at each level in the ultimate hierarchy. ([p.] 4/MAXTHINK) The Library View of Hypertext Except for a few general articles, the library literature has essentially ignored hypertext system activity and promise. There has not been library appreciation or application of the technology for substantial information access system purposes. The university libraries have been involved in the major Brown University and USC hypertext projects mentioned above, but they are portrayed primarily as cooperative team members within the respective academic communities. They do not appear to have been real innovators or leaders of these projects. Literature on these two projects addresses primarily the educational and multimedia applications, not the nitty-gritty concerns of knowledge organization and information access development. Bevilacqua (1989) <refs -bevil>, Debuse (1988) <refs -debuse>, and Franklin (1988) <refs -franklin> have contributed short introductory articles in the library press. They give general overviews of the hypertext associative linking approach, and suggest possible impact on libraries and information retrieval. Gaines and Vickers (1988) <refs -gaines> have produced a more substantial piece, delving into the theoretical constructs and possibilities of hypertext and hypermedia systems. Urr (1991) <refs 23 8> discusses the lack of organization and structure evidenced or supported by most reported hypertext implementations. He echoes Gaines and Vickers in criticizing false or pseudo-hypertext system claims. The major notice of hypertext potential by the library and information community has taken place at Linköping university in Sweden, by investigators associated with the LIBLAB project. Hjerppe and his associates have undertaken this substantial research effort. The LIBLAB investigation is studying the effects and potentials of the evolving electronic information environment upon the library of the future. A primary focus of the research program involves development of the HYPERCATalog system. This is envisioned as a hypertext format online public access catalog (OPAC), as well as a personalized information system for library users. Hjerppe has published numerous outlines of LIBLAB project goals and plans (Hjerppe, 1990b <refs 13 13>, 1989 <refs 13 4>, 1986 <refs 12 20>; Hjerppe, Bivens-Noerr, and Noerr 1982) <refs 13 24>. He has also presented detailed considerations of knowledge organization and the evolution of the library roles in the emerging electronic society (1990a <refs 13 8>, 1990c <refs 13 19>) Björklund describes the possibilities of the HYPERCATalog as a researcher tool (1990a) <refs 2 21>. She has also reported on the experimental hypertext conversion of a section of the Swedish "SAB" hierarchical classification (1990b) <refs 3 4>. As in other LIBLAB electronic classification and cataloging code efforts, they found that hypertext conversion of the hierarchical classification highlighted inconsistencies in the existing classification. Vainio-Larsson, another LIBLAB researcher, has produced two articles (1989a <refs 23 18>, 1989b <refs 24 1>) delving into hypertext system/user interactions and the potential HYPERCATalog application. He analyzes the concept of the user mental model of the system, and the role of metaphor in the system interface. Literature Review Conclusion This review of the general information access literature and specific hypertext literature provides a base for the present study. In this review, the writer has shown the background and need for the study. This study is an analysis of traditional information access approaches and their possible application to a new electronic information delivery medium.