Resource Library: Multimedia

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

/ Resource Library: Multimedia / Resource Library: Multimedia.iso / hypertxt / msdos / hypdiss / app-e < prev next >

Wrap

Text File | 1992-04-02 | 43.4 KB | 865 lines

APPENDIX E, HYPERTEXT INFORMATION ACCESS STUDY INTERVIEW NOTES TONY PHILLIPS INTERVIEW KENTFIELD, CAL. MARCH 15, 1991 A. HYPERTEXT ARCHIVE TRANSACTION/SUPPORT SYSTEM: A.1. Please summarize the basic hypertext document content, assembly & maintenance procedures? MaxThink, Inc., Berkeley CA; and Pondlife, Inc., Kentfield CA; produce the DaTa database, under contract to Deloitte & Touche. Anthony Phillips is primary author for the DaTa CD-ROM. Content - The database consists of both source text and bibliographic reference records relating to American, Canadian, and international auditing and accounting practice. The input text is selected by Deloitte & Touche professionals, and sent to Phillips almost entirely in hard copy print format. Phillips estimates that he receives no more than 5% of the material in electronic format, suitable for direct processing. Database size - He estimates the database currently consists of some 20,000-25,000 files, maintained in a hypertext network web, which incorporates some 200 sub-networks. The database takes some 80 megabytes of disk space; although there are only approximately 52 megabytes of actual text file or "logical" records; the additional disk space is consumed because of the nature of MS-DOS disk sectoring. Phillips adds or updates approximately 3000-4000 documents per year. This results in addition of about 6 megabytes per year. He estimates 15% of the database "dies" by replacement or deletion each year. Conversion - Since most of the information is in hardcopy form, the data must first be converted to ASCII text in MS-DOS format, for entry into the hypertext system. They use a Kurzweil Discover 7320 OCR unit, a top-of-the-line OCR product, for the conversion. They elected to use MS-DOS text format, for compatibility with the majority of microcomputers used by field accountants and auditors. The conversion is labor-intensive and therefore expensive: Phillips estimates that conversion and reformatting accounts for 80% of the total production effort. Reformatting need - He points out that ALL the text must be translated into an ASCII format suitable for their simple display. He says, "What looks good in print sure as hell doesn't look good on the screen! We must do a lot of beautifying." Phillips states that the extensive reformatting is necessary, since their system must work with the lowest common denominator of end user screen display. They plan or allow for 24 lines, straight or plain ASCII text display, and for an unknown quality of monitor. Reformatting process - Phillips says there are three basic steps in preparing the text for hypertext format: 1) "Get it clean." - This refers to spellchecking and source proofing of the text produced by the OCR conversion. 2) "Chunk it." - He goes through the text, to mark it for splitting into small topical units of nodes, "into lumps that make some sense." He inserts "flags," or markers for automatic processing with the "File Splitter" utility (below). 3) "Split it." - Phillips uses MaxThink's FS (File Splitter) utility. This program reads the original text file. It identifies flagged sections and writes them to separate ASCII files. It also produces a worklist file, containing the file names, and comments,which were embedded in the identifying flags. Updating task - The hypertext textbase has been assembled over the last several years. As mentioned above, Phillips now estimates that approximately 15% of the material becomes obsolete each year. He must editorially decide between complete deletion of this material, or whether to leave it for archival record, with indication of the deleted or superseded status. Phillips favors the archival approach, for material of permanent reference value. He feels that much of this will be important to potential researcher or university users. "Besides," he says, "it cost us an arm and a leg to get [convert] the stuff! Why throw it away?" A.2. GENERAL PRESENTATION AND PRODUCTION DESIGN OF THE HYPERTEXT ACCESS SYSTEM: A.2.a. Describe the general arrangement of the main document file. (Unique document identification, general logical arrangement, basic principle of access) The documents are simple ASCII files, stored in a series of disk subdirectories. Present and earlier generations of MS-DOS can not efficiently store a large number of files in a single disk directory. Overloading a system directory results in poor file retrieval and display performance. Therefore, Phillips sets a limit of about 100 files per directory, to avoid this degradation. Even with this requirement for artificially splitting the files into small subdirectory batches, the MaxThink software can locate and access files fairly easily. When the software searches for a filename, it automatically searches for a disk subdirectory matching the first 3 characters of the given filename. For example, if searching for a file named "CODE37A.TXT", the software will automatically check for an MS-DOS subdirectory named \COD. If it is present, it will find the file in that directory. This default check makes it efficient to locate and retrieve files stored in the limited small groups. Phillips uses a combination of methods for assignment of files to specific subdirectories. Some are defined using a simple classification principle, such as issuing agency, journal article source, etc. He assigns other documents based on a personal conceptual classification system. To do this, he analyzes the document content according to 1) Entity, or type of organization; 2) Subsystem within the entity; and 3) Transaction type. [In effect, he is using a limited form of facet analysis of the various aspects of the document content - INVESTIGATOR] As described above, MaxThink software efficiency depends upon the first three letters of the filename duplicating the three-letters subdirectory code. After he decides upon a specific subdirectory storage location, Phillips then uses a variety of approaches to form the rest of the unique MS-DOS filename. If the document is issued by an agency or promulgating body, he often uses that agency's document or code number. He may optionally use his standard syntax for identifying the document section, subsection, page number, etc. A.2.b. Please summarize the general concepts of the system's "user interface," the document access and display methods, design of the presentation means, etc. Software interface: Phillips says the system was designed using a "minimal expectation of user ability." He calls it a "one- finger system." After entering the system, the user can traverse the hierarchy and the network connection, and display document content by simply using the four cursor arrow keys. There are, of course, other operations handled by a few function and mnemonic alphabetic keys, such as "P" to print a document, "V" to view the hierarchy trail leading to present position, F1 to display Online Help, etc. Phillips reports that they attempted to "minimize the mental load for using the system." He says, "We wanted the information to be the issue, not the software." "We knew we were on the right tract when we see new users begin interacting with the CONTENT, agreeing or disagreeing with what they see on the screen, within a few minutes of starting to use the product. They totally forget about the system." He feels that other hypertext system designers have let software technology sophistication interfere with simplicity and ease of operation. "I have a lot of trouble with HyperCard," he says, "because of the plurality of commands. And, then, when you get to something like OWL Guide, KnowledgePro, and the rest.... The issue's got to be the information, not the software!" Conceptual metaphor: As a domain expert, Phillips has developed a conceptual matrix covering the accounting/auditing knowledge domain. [He is well qualified in this area, with international practice experience with the now- absorbed Touche-Ross firm; and service as their representative to the Financial Accounting Standards Board, a major industry policy body.] He has literally translated the detailed conceptual matrix into a interconnected web of over 200 hierarchical networks. He uses network-building tools, described below, to build and maintain control over the intricate knowledge representation. The software lets him build a complex, three-dimensional, conceptual web. But at any one point, the user sees only a flat, two- dimensional screen representation of exclusive topical choices. The system thus retains the hierarchy's advantage of offering clear, simple, choices to the user, at every point. As Phillips describes it, "Every screen represents a decision on how to reach his answer, to find his topic." He says their system provides the power to use descriptive metaphors, which can communicate to the user the likely content behind the various screen choices. Before working on document access, he says that he must first identify the multiple substantive or "commonsense issues" covered in the document. Like more traditional library catalogers or indexers, he observes that he must do this "without knowing the user's question." Phillips says that after he positions the document file in the proper storage area, as described above, he must then set about providing multiple access points to the document. In analyzing a document, Phillips says that he "goes to the bottom of the matrix, and looks up." He then connects the document to the network trails leading "down" to the pertinent topical areas and nodes. A.2.c. Identify and briefly describe the general production tools or building tools used in construction of the system. They use a variety of tools to create the hypertext system: 1) "Cleaning and chunking tools" - Turbo Lightning (spell checker) to remove data errors, document-splitting utilities to divide the documents into conceptual nodes, ordinarily a screen or two in length. 2) Pre-hypertexting tools - TR (Word processor/Editor), to reformat the document text for best screen presentation. It is also used for embedding or insertion of hypertext links within directly into text of information base documents. 3) Network-building tools - Phillips uses Houdini, a matrix-outliner as the main tool for building and updating the complex hypertext hierarchy and network. He uses Houdini to maintain a "3-dimensional" conceptual map outline if the domain. He is able to use the matrix outliner locate or place documents into the network. He describes this as "creating the conceptual map." Phillips simultaneously uses two networked AT-class computers for the hypertext authoring task. There are also two other microcomputers running, for allied support tasks. The two networked units run Carousel multi-tasking software. Each of these machines can simultaneously access the current CD-ROM database, as well as the "new material" on magnetic disk. The two networked computers each simultaneously operate multiple task copies of a DOS shell utility, the matrix outliner, the CD-ROM software, the TR editor, and other useful utilities. The end result is that each computer is running simultaneous copies of multiple authoring tools, as well as being networked together. The networking allows easy file transfer, as well as sharing of the CD-ROM and other peripheral devices. This collection of software tools makes it possible for Phillips to handle multiple tasks, as necessary. He begins with the "current" CD-ROM master network as a base for the future network. He then modifies it to include the new and updated material, and also flags files for deletion. He is thus constantly updating and modifying the definitions of the multiple included networks. As noted, Phillips has multiple sessions running on each machine, with a variety of software tools at hand. He may use simultaneous, windowed display of a new document in the editor; related older CD-ROM documents loaded into the editor for added linking or modification; Houdini display of the different complex and interlocking networks; a copy of Houdini where he is inputting phrases for the KWOC index; file transfer utilities, etc. For example, during the simple interview demonstration of methods to this investigator, he had three or four sessions running on each machine. Phillips commented that he normally has a total of ten or twelve sessions simultaneously running, on the two network-building machines. Phillips also uses the "TransText" editor to insert embedded links directly into document text. These include hypertext links to topically related files, to network nodes, and linking sequentially preceding and following text files, for documents which have been split up, or "decomposed." Phillips feels that the hypertext links embedded in text are obviously important. But he feels that the conceptual hierarchical network is the major intellectual and access value contribution. The network links offer multiple access paths to a given topic, much like a multiple-entry library catalog. The major advantage of hypertext is that the source or text material is instantly available for quick retrieval and display at the "end" of the network trails. This is different from older information retrieval models using final references to remotely stored document media. MaxThink CD-ROM production methodology has evolved to the point where their CD service bureau maintains the master working-file tape copies of the "current state" of the CD-ROM. Phillips now submits only update "changes" to the current database. These changes typically include: 1) new text and network files; 2) text and network files to be updated (after modification); and, 3) files to be deleted. The service bureau has developed utilities to merge these changes into the current version, and can then proceed with production of the new CD-ROM. A.2.d. Identify and briefly describe the specialized organizational and quality control tools which allow you to build the system. Houdini network builder - Phillips feels the most important hypertext-building tool is the Houdini matrix outliner or network-builder program. He does not use the MaxThink hierarchical [flat] outliner, but only the Houdini program. He uses it to build complex conceptual networks of domain topics and descriptors. These networks are also interconnected. The software allow any network node to link to any other network top level node or internal node. Houdini handles all the necessary network links, and can also display the text of the final referenced document files. Phillips points out that building a complex conceptual network like this absolutely requires a powerful and sophisticated matrix tool like Houdini. KWOC index - Houdini also contains a "glossary- building" tool. This facility allows the author to selectively flag network topics and descriptive file pointers for inclusion in an online index. This is an online KWOC index to all flagged network topic levels and documents. The online index can be easily displayed from virtually any point in the hypertext system. KWOC entries themselves include hypertext links, affording fast access to any included network level or document node. The ease of operation makes it an effective and economical indexing mechanism. Hypertext editor - Phillips uses the TransText editor to embed or insert interdocument links into the text of document nodes. This editor can also execute hypertext jumps, so it is an effective authoring tool. He uses it to move throughout the network of text documents, inserting links, making on-the-fly changes, etc. But he points out that this work generally involves working with only a limited cluster or area of topically pertinent document files. Any ASCII editor with multi-document windows is capable of managing such a micro-collection of information. But he feels that this brute force "editor-only" approach can simply not effectively manage the tens of thousands of documents in a mega-hypertext. He reiterates that a network manager is absolutely essential for effective hypertext authoring. Production tools - There are assorted other tools used for efficient translation of the network into a CD-ROM format. These are strictly processing tools, and carry no real significance in system creation or conceptual design. B. THE HYPERTEXT INFORMATION ACCESS SYSTEM: B.1. ACCESS POINTS - Which of the following types of access points are included in your system? For each question item, please rate using the following categories, and comments as needed... P)resent,E)asily achievable,M)odifications needed,N)ot achievable B.1.a. Main file sequence - direct file access Category: [P] E M N The text files are divided into MS-DOS subdirectories, containing a maximum of about 100-125 files, to maintain DOS efficiency. Within the system, a file may be directly accessed with a simple "GOTO" command, using the convention of the DOS filename within angle brackets, for example, "<TESTFILE>". B.1.b. Author Category: [P] E M N MaxThink uses author access within some areas of their information base. Author access may easily be provided by building links from author listings. This is an editorial option. As Phillips phrases it, "Representation form decision is simply a choice of the best way to represent the information." (This applies to most of the "editorial option" answers below.) B.1.c. Title Category: [P] E M N Author access may easily be provided by building links from author listings. This is an editorial option. MaxThink uses author access within some areas of their information base. B.1.d. Name forms Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. B.1.d.i. Personal names Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. B.1.d.ii. Corporate names (Companies, organizations, government, etc.) Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. B.1.e. Keywords Category: [P] E M N The "Glossary" module of the MaxThink software gives KWOC access to both network and file nodes. B.1.f. Subject/Topic/Concept Category: [P] E M N The network access design is based upon end user choice of topical or subject coverage. The author describes the network subject choices as "subject metaphors," or "creative descriptions." He attempts to coin these descriptive headings or choices as three- word phrases, per Lakoff. B.1.g. Geographic Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. Phillips notes that it would appear to be a domain- specific value. If geographical access is implemented, he would recommend providing it with Houdini network input to create Glossary KWOC entries. B.1.h. Date, chronological, temporal Category: P [E] M N DaTa does not consistently use this access approach, but it is an editorial option. Phillips notes that he occasionally does provide date access for promulgating agency documents. B.1.i. Language Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. B.1.j. Document format - book, article, pamphlet, report, etc. Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. Their file-naming syntax partially provides this kind of information, but Phillips would recommends using Houdini network links to represent it. B.1.k. Document position - section, page, location Category: [P] E M N This is an editorial representation decision. The menus or document nodes can reflect and provide access to specific pages, locations, document sections, etc. The MaxThink hypertext runtime software also provides string search access within a text file, and can be used to locate regular heading or pagination forms, etc. This would not be equivalent to postcoordinate, indexed field, search, but could provide the same kind of access. B.1.l. Automated field specifications - record size, entry date, notations, originator, etc. Category: P [E] M N The MaxThink hypertext software cannot itself do any kind of Boolean search of the information base. It DOES have the ability to use links or function keys to invoke external programs. This can include the ability to run database or text-searching programs on the same (or a surrogate) information base. B.2 ACCESS APPROACHES - Which of the following subject or topical information devices are used in your system? For each question item, please rate using the following categories, and comments as needed... P)resent,E)asily achievable,M)odifications needed,N)ot achievable B.2.a. Classification schemes (NOTE: On opening discussion of this topic area, Phillips observed amusedly that this area of classification schemes was an "entirely subjective" one. As he put it, "We're pointing to the same box of ASCII files. If you disagree with my classification, use your own, or use a combination of multiple access structures.") B.2.a.i. Hierarchical taxonomy Category: [P] E M N The entire basis of the MaxThink approach to hypertext is the use of controlled hierarchical taxonomies, especially in interconnected network form. B.2.a.ii. Enumerative, universal, classification [Dewey type classification] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. This would be easily implemented by creating or importing the classification file with the MaxThink outliner. (Phillips describes this tool as the one to use for "hard hierarchies." The basic classification could then be moved into the Houdini matrix tool, for easy linking to the information base text files. B.2.a.iii. Specialized, literary warrant, classification [Library of Congress, Reader Interest Classification] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author could use the same approach for classification operations as in the preceding item. B.2.a.iv. Faceted classification (analytico-synthetic) [PRECIS style of indexing] [C., p.65] Category: P [E] M N This is an editorial representation decision. In this case the author would need to use external software to build and maintain the faceted classification representation. The hypertext authoring software could easily import ASCII text file output produced by the external software, and use them in building links. B.2.b. Indexing approaches B.2.b.i. Alphabetical index, separate or dictionary file Category: [P] E M N This is an editorial representation decision, an alphabetical index can be manually produced with the present authoring tools. The hypertext software can also utilize ASCII text output from standard indexing software. B.2.b.i.A. Keywords, extracted or assigned Category: [P] E M N The DaTa CD is produced including a KWOC index, produced from assigned descriptive phrases, and descriptive text lines extracted from information base text files. External software utilities could also be used for keyword extraction. B.2.b.i.B. Controlled vocabulary assignment Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author can use manual means, or external software, for maintenance of the controlled vocabulary. B.2.b.i.C Relative index, e.g., to Dewey classification Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author can use manual indexing means, or external software, for maintenance of the index. B.2.b.ii. Term manipulation indexes (generally for production of printed output) Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. This would require an external software utility, to produce index in ASCII text format, for hypertext linking source. B.2.b.ii.A. Simple permuted or rotated - KWIC Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. This would require an external software utility, to produce index in ASCII text format. B.2.b.ii.B. Ordered by extracted element - KWOC Category: [P] E M N The present DaTa CD contains an online KWOC index with hypertext links to network and file nodes. B.2.b.ii.C. String indexing (phrase-manipulation, rotation of terms) - PRECIS, NEPHIS, etc. Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. This would require an external software utility, to produce index in ASCII text format. B.2.b.ii.D. Chain indexing (string indexing, with forms reflecting basic taxonomy of terms [C., p. 67] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. This would require an external software utility, to produce index in ASCII text format. B.2.b.iii. Classified index (generally requires secondary alphabetical index, for ease of use) [C., p. 56] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author can use manual means, or external software, for maintenance of the controlled vocabulary. B.2.b.iv. Coordinate indexing - Manual coordination or automated database file, using Boolean search [C., p. 60] Category: P [E] M N Easily be accomplished by hypertext link call of external text or database searching program. B.2.b.iv.A. Older non-automated searching methods - peekaboo, edge-notched cards, Uniterm terminal digit cards Category: P E M [N] Not applicable to electronic system. B.2.b.iv.B. Database file search - Sequential or indexed field search Category: P [E] M N Easily be accomplished by hypertext link call of external text or database searching program. B.2.b.iv.C. Full text search Category: P [E] M N Easily be accomplished by hypertext link call of external text or database searching program. B.2.b.v. Faceted indexing [C., p 65] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author will need external software, to generate and maintain the faceted index. B.2.b.vi. Citation indexing [C., p. 72] Category: P [E] M N DaTa does not presently use this access approach, but it is an editorial option. The author can use manual means, or external software, for maintenance of the citation index. A citation index using hypertext links could display cited electronic text, or could link to external systems for presentation of more information. This could include online databases, online library catalogs, etc. B.3. CONTROL MECHANISMS - Which of the following subject access control measures, intended to control consistency, form, and item sequencing, are present in your system? For each question item, please rate using the following categories, and comments as needed... P)resent,E)asily achievable,M)odifications needed,N)ot achievable B.3.a. Classification schedule Category: [P] E M N The domain-specific taxonomy created by Phillips is essentially a "classification schedule," in expanded, networked format. The matrix outliner affords great flexibility in modifying the "classification" form. It also simplifies relocating the linked text nodes. B.3.b. Vocabulary control systems Category: P [E] M N This is an editorial decision. Vocabulary control can be accomplished using manual or external software means. B.3.b.i. Authority/Headings files Category: P [E] M N This is an editorial decision. Vocabulary control can be accomplished using manual or external software means. B.3.b.ii. Thesaurus control Category: P [E] M N This is an editorial decision. Vocabulary control can be accomplished using manual or external software means. Simple synonym control is already taken care of in the KWOC index, using a utility program. B.3.b.iii. Derived-term methods or algorithms Category: P E [M] N Possible, using external program, to generate text for import into the hypertext format. Phillips observes, "If you're a domain expert, you don't need it." (Referring to the person authoring the hypertext.) B.3.b.iv. Hierarchical search thesaurus (for database file search) Category: P E M [N] Not possible or useful with the hypertext associative linking approach. Could certainly be a feature of external searching software called up by link commands. B.3.b.v. Entry term form control mechanisms Category: P [E] M N This is an editorial decision. Entry form control can be accomplished using manual or external software means. B.3.b.v.A. Entry syntax (preferred noun/adjective, etc., construction form) Category: [P] E M N Present editorial practice for entry of network descriptive phrases. Can easily use external software for greater control, if deemed necessary. B.3.b.v.B. Standard number approach (plural, singular form preference) Category: [P] E M N Present editorial practice for entry of network descriptive phrases. Phillips uses software utility to depluralize and clean up glossary index listing. B.3.b.v.C. Automatic depluralization (database file) Category: P [E] M N Not possible or useful with the hypertext associative linking approach. Could be a feature of external search software called up by link commands. Automatic depluralization IS used within the MaxThink KWOC indexing. B.3.b.v.D. Synonym definition (database file) Category: P [E] M N Not at all possible or useful with the hypertext associative linking approach. Could be a feature of external search programs called up by link commands. Synonym control IS used within the MaxThink KWOC indexing module. B.3.c. "Standard Subdivision" or faceted classification protocol Category: [P] E M N Used in present editorial approach, for automatic ordering of network entry types in the glossary index KWOC file. B.3.d. Term or descriptor relationships - Roles, links, weighting Category: P E [M] N Not currently implemented. Have considered this possibility, as a "filtration" or "selective viewing" device to help the end user. Already have "hooks" in the software which could use this approach to display only those links meeting assigned or selected features, instead of all embedded links. B.3.e. Filing or sorting rules Category: [P] E M N Presently use straight ASCII sort in glossary index listing. Could alternatively use external utility and use some other sorting algorithm. B.3.f. Manual or automated authority/procedural safety measures Category: [P] E M N Use multiple backup approaches. FILE BACKUP - These files contain copies of taxonomies, multiple networks, index source text and KWOC listing files. 1) Archive of past CD-ROMS, both off- and on- site. 2) Daily streaming-tape backup of active authoring file. PAPER BACKUP - Hard copy of the network hierarchies. Produced by printing copy of the network matrix with the MaxThink flat outliner software. This gives multiple-entry listing of the interlinked network files.