Access to Networked Documents

Contents

• Study
• References
• Figures

In a study completed in December 1995, we were trying to identify the metadata in use for retrieving electronically-stored information on the networks. We came to the overwhelming conclusion that in order to do an effective search, one must understand the search engine being used, its database size and contents, and its underlying index strategy. Actually, this has always been true, even in the use of print indexes and catalogs. For example, if a user approaches a traditional catalog with a citation to a journal article and attempts to find the author and title of the article, the result will be "no hit." This happens because the user does not understand that the contents of a database called a "catalog" does not include titles of individual articles from journals, but only the titles of the journals themselves. In the same way, each search engine has its own limits on contents and its own set of elements that it will search. In addition, each has its own way of displaying results. The problem is that it is very difficult to determine just what is included. Several search engine beginning pages proudly display "how many," but not "of what." When one does the same search on several search engines, one quickly learns that

• No single indexing service can reach everything on theInternet.

  and

• Searches produce mixed and unpredictable results.

Internet search engines have both strengths and weaknesses. Among the strengths are:

• One can access these indexes 24 hours a day, seven days a week.

• Each returns a list of hits, each linked to "document-like objects" which are usually available immediately.

• The Internet resource pool is dynamic--growing all the time.

• Personal pages and other ephemera are accessible without requiring intervening selection, processing, and cataloging decisions.

• They are multidimensional--one can get to library catalogs and other search engines from most of these indexes.

Among the weaknesses are:

• There is much duplication of entries within the same set of retrieved "hits."

• Results are unpredictable.

• Results can be quite misleading--the same search can retrieve no hits by one engine and many hits by another.

• Search engines do not readily disclose the contents of their databases nor do they provide a description of the criteria used to include a document in their files.

• Vocabulary is not controlled, and punctuation and capitalization rules are not standardized. (The illustrative figures are from "Alta Vista," the widely touted new search engine of the Digital Equipment Corporation.) E.g.,
  • "SAD" for seasonal affective disorder produces many references to the feeling (predominately lonely-hearts and personals pages). In a search for "SAD" in Alta Vista, the first relevant hit is number 11 (see fig. 1). Using uppercase in the search does not help (see fig. 2), and using periods after the letters causes other difficulties (e.g., among other things "S.A.D." retrieves "Students Against Drunk Driving").

  • Another problem is the combination of common words such as "Data administration." When adjacent and in the proper order, they describe the administration of electronic data objects. Otherwise, the terms can retrieve a large body of information about data processing as it relates to business administration.

  • Names of persons are particularly difficult to find. A search for arlene taylor, for example, retrieves approximately 10,000 hits. The first 20 are essentially duplicates (i.e., pieces of the same home page) (see fig. 3). The 21st to 29th hits are totally irrelevant if one is looking for someone named "Arlene Taylor" (see fig. 4). The 30th hit, however, contains a bibliography that has a listing including that name. It should be mentioned that putting the name in quotation marks in the search will retrieve only items where the two names are together, but it will miss the home page for "Arlene G. Taylor." Putting in the "G." will get the home page but will miss the faculty listing for the School, where the "G." is not included. And both listings will miss bibliography entries where "Taylor, A.G." is used.

• Relationships and relevance often cannot be analyzed withoutactually examining each item--that is, there is not enough information in the "index entry" to allow one to make educated choices. For the 30th hit from the arlene taylor search above, for example, one must search the entire document to determine the reason for its retrieval (see fig. 5). And many search engines do not even give the amount of information shown in figure 5. Instead retrievals look much like Alta Vista's "compact listing" (see fig. 6)

Understanding these indexes and their respective search features in combination with our own knowledge of information-seeking can improve the quality of our hits. The search for information on data administration is a good example. An initial search on these two terms will return any document containing one or both of these terms. Even in queries conducted on search engines that weight hits according to adjacency of terms this combination can produce too many irrelevant results. A second search on the same terms, enclosed in quotes (if the search engine provides this sort of qualifying device), will produce a list that is smaller and more precise.

Over the last few centuries since the last information explosion crisis (i.e., the invention of printing), we in the library and bibliography field have developed principles of bibliographic control that have proved quite successful in aiding us to organize information so that it can be found again. One of these is that we should be able to design a citation that presents an integrated and intelligible description of a document (see fig. 7). This citation should clearly indicate the document's relationship to its own earlier (or other) manifestations as well as its relationship to other documents in the database. In other words the purpose is to provide enough information to enable the searcher to choose what is going to be relevant and useful to that searcher (see fig. 8). A citation does this by providing:

• Name of the document

• Names of those responsible for the intellectual contents

• Edition or issue information

• Place of creation (for purposes of identification of such things as point of view, including such things as geographic location and/or name of institution)

• Age information (e.g., date of creation of intellectual contents)

• Pertinent physical features that may affect choice (e.g., size in bytes, or any special devices necessary to view the document)

• Relationship to other entities (e.g., its number in a set of documents; or the print document from which this is copied)

• Subject descriptors in controlled form

• Location of the document

• Other information that any specific document may need to have included in order to provide an intelligible description

This is the principle underlying many of our current indexing services (e.g., the H. W. Wilson indexes), as well as library catalogs. AACR2 is a strong standard for description, although it is focused primarily on physical documents; however, one can follow the spirit of AACR2 without necessarily following every rule to the "letter."

We have focused on the issue of cataloging the whole Internet from the outside in. In fact, catalogers have the opportunity to affect the quality of Internet publication from the inside out. We can

• work within our own organizations to set standards for home page development

• network with our peer institutions to develop other standards which are useful to the community

• urge our local community members to provide keywords and other descriptive information with their homepages

• provide other means for enhancing access points in locally produced documents.

We would like to suggest that this is the underlying principle being put to use in creation of what Priscilla Caplan has called "syntax-independent metadata for document-like objects." She identifies "syntax-independent" as meaning that "the data may be represented in USMARC, HTML, SGML, or 'Keyword3Dvalue' pairs." (p. 21) We should recognize that it is the principle developed especially over the last 150 years that we want to defend and to insist be used by organizers of the Internet, not an undying affection for AACR2 per se. Standards of description assure built-in predictability and eliminate most duplication by analyzing the linkages between entities. There are documents of lasting value among the contents of the Internet's amorphous database.These are the documents to which the principle should be applied. Organizers should select the vehicle of choice (MARC, TEI, HTML, etc.); and do it!

REFERENCE

Caplan, Priscilla. "You Call It Corn, We Call It Syntax-Independent Metadata for Document-Like Objects." The Public-Access Computer Systems Review 6, no. 4 (1995): 19-23; also found at:

http://www.nlc-bnc.ca/documents/libraries/cataloging/caplan3.txt

Documents 11-20 of about 2000 matching some of the query terms, best matches first.

Mental health links

• Links on SAD, depression and mental health. These are bookmarks I have saved whilst exploring search sites and indexes. I have not had time to read many of them yet myself, so some may not be relevant and I can not give any details on the sites. For...

http:01/30/96/www.geopages.com/Hollywood/1872/jumpsad.html -size 4K - 19 Oct 95

Documents 1-10 of about 2000 matching some of the query terms, best matches first.

No Title

8 - Problemas com UTM/SAD-69 Submetido por: Eduardo Burla FURNAS FURNAS CENTRAIS ELETRICAS Data de submissão: Quarta-feira, 5 de Abril de 1995, 15:19 Máquina remota: maestro.dpi.inpe.br (150.163.2.249) Produtos afetados: spring Descrição do problema ou...

http:01/30/96/www.inpe.br/spring/bugs/bug.8 - size 454 bytes- 5 Apr 95

No Title

SAD DAY IN TAMPA BAY. MICK IN VALRICO (XFRJ11B@PRODIGY.COM) Saturday, 18-Nov-95 12:15:11 CST. Well we're 18k short, hope everyone will be at the flea market Sunday cuz that is the future- pathetic showing of support and pride for your city-residents and...

http:01/30/96/www.hamline.edu/~sfitchet/bucs/buctalk/256.html - size 526 bytes - 13 Dec 95

Fig. 3:

Arlene G. Taylor Home Page. Associate Professor Department of Library Science School of Library and Information Science 642 SLIS Building 135 Bellefield Avenue Pittsburgh, PA 15260. TEL: (412) 624-9452 FAX: (412) 648-7001. Curriculum Vitae. Course Syllabi....

http:01/30/96/www2.pitt.edu/~agtaylor/ - size 681 bytes - 1 Dec 95

(Hits 1-19 were sections of this page.)

All My Children Update for Thursday September 7, 1995

Update for AMC Thursday September 7, 1995. By Ashley Lambert-Maberl y. Welcome to the Thursday update! SPERM WAILS. In which Taylor wonders whether Dreck's magic sperm did the trick ... in which Arlene wants Alec to give her more of his special stuff ...in...

http:01/30/96/purplenet.com/soaps/amc/stories/sep95/sep07r.html - size 7K - 13 Sep 95

No title

Source: This information is provided by Rebecca Green, College of Library and Information Services, last revised January, 1994. For further information contact Dr. Green at (301) 405-2050 or electronic mail at rgreen@umd5.umd.edu LBSC 671 Rebecca Green...

http:01/30/96/www.inform.umd.edu:8080/Educational_Resources/CollegesAnd Schools/CLIS /Courses-Spring94/LBSC671-0101/readings

• - size 5K - 12 Dec 95

Documents 21-30 of about 10000 matching some of the query terms, best matches first.

No Title 090909[21Apr95]                  Date: Sat, 15 Apr 1995 16:54:5
Arlene Coddington 0909[22Nov95]           Arlene Coddington. REALTORAE As
Home Page: Arlene Badman  09[18Jul95]     Jane C. Taylor, M.F.A. Jane re
management_team.html 0909[27Oct95]        SIS2000 Project Management Tea
Student Affairs Projects  09[11Dec95]     Projects - Major Organization.
List of Professional Deve 09[24May95]     The Professional Development M
All My Children Update fo 09[13Sep95]     Update for AMC Thursday Septem
Titles OPQ 090909[08Dec95]                D o c u m e n t s b y T i t l
NUKE Moviescape Reference 09[04Dec95]     Rebecca De Mornay. Filmography
No Title 090909[12Dec95]                  Source: This information is pr

To design a citation that :

• presents an integrated and intelligible description of the document
• clearly indicates the document's relationship to it's own earlier (or other) manifestations
• clearly indicates the document's relationships to other documents in the database

To provide enough information to enable a searcher to choose what is going to be relevant and useful to that searcher

Back to Top