Search (219 results, page 3 of 11)

0.0028703054 = product of:
  0.005740611 = sum of:
    0.005740611 = product of:
      0.011481222 = sum of:
        0.011481222 = weight(_text_:a in 400) [ClassicSimilarity], result of:
          0.011481222 = score(doc=400,freq=4.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.2161963 = fieldWeight in 400, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.09375 = fieldNorm(doc=400)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.0028703054 = product of:
  0.005740611 = sum of:
    0.005740611 = product of:
      0.011481222 = sum of:
        0.011481222 = weight(_text_:a in 2471) [ClassicSimilarity], result of:
          0.011481222 = score(doc=2471,freq=16.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.2161963 = fieldWeight in 2471, product of:
              4.0 = tf(freq=16.0), with freq of:
                16.0 = termFreq=16.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046875 = fieldNorm(doc=2471)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

This is a short, but very thorough and very interesting, report on how the writers built a faceted classification for some legal information and used it to structure a web site with navigation and searching. There is a good summary of why facets work well and how they fit into bibliographic control in general. The last section is about their implementation of a web site for the Washington State Bar Association's Council for Legal Public Education. Their classification uses three facets: Purpose (the general aim of the document, e.g. Resources for K-12 Teachers), Topic (the subject of the document), and Type (the legal format of the document). See Example Web Sites, below, for a discussion of the site and a problem with its design.

Content

A very large PDF of the six-foot-wide illustrated poster from their poster session is available at http://depts.washington.edu/pettt/presentations/conf_2003/IASummit-Poster-Louie.pdf.

0.0028047764 = product of:
  0.005609553 = sum of:
    0.005609553 = product of:
      0.011219106 = sum of:
        0.011219106 = weight(_text_:a in 7410) [ClassicSimilarity], result of:
          0.011219106 = score(doc=7410,freq=22.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.21126054 = fieldWeight in 7410, product of:
              4.690416 = tf(freq=22.0), with freq of:
                22.0 = termFreq=22.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0390625 = fieldNorm(doc=7410)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Once the library has been through a process of transition from traditional library procedures to automated ones, natural language searching became a necessity for both indexers and searchers. Therefore, aside from the precoordinated classified catalogue we started to build a dictionary of terms in order to make postcoordinate search possible in keeping with the UDC notations assigned to each bibliographic record. After a while we came to the conclusion that the dictionary needed a control of its terms so that synonymous concepts and semantic ambuguities be avoided. The project presented in this paper shows how reality imposed the improvement of the quality of indexing and hence of the searching possibilities. Is also shows the reasons why we consider a multilingual thesaurus based on UDC an ideal indexing and searching device. The experiment applied on class 8 of UDC illustrates the way the UDC tables can be quite successfully used in building a thesaurus due to their qulities and how their limitations can be overcome by a thesaurus. An appendix to the paper contains a sample of the multilingual thesaurus given in both alphabetical and systematic layouts

Type

a

0.0028047764 = product of:
  0.005609553 = sum of:
    0.005609553 = product of:
      0.011219106 = sum of:
        0.011219106 = weight(_text_:a in 2480) [ClassicSimilarity], result of:
          0.011219106 = score(doc=2480,freq=22.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.21126054 = fieldWeight in 2480, product of:
              4.690416 = tf(freq=22.0), with freq of:
                22.0 = termFreq=22.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0390625 = fieldNorm(doc=2480)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Three pre-web articles about using hypertext for knowledge representation. Duncan discusses how to use graphical, hypertext displays (she used Xerox PARC's NoteCards on a Xerox 1186 workstation) along with concept maps and facet analysis, a combination that would now be done with topic maps. The screen shots of her graphical displays are quite interesting. Her interest in facets is in how to use them to show things to different people in different ways, for example, so that experts can enter knowledge into a system in one way while novices can see it in another. Duncan found that facet labels (e.g. Process and Product) prompted the expert to think of related concepts when inputting data, and made navigation easier for users. Facets can be joined together, e.g. "Agents (causing) Process," leading to a "reasoning system." She is especially interested in how to show relstionships between two things: e.g., A causes B, A uses B, A occurs in B. This is an important question in facet theory, but probably not worth worrying about in a small online classification where the relations are fixed and obvious. These articles may be difficult to find, in which case the reader can find a nice sumary in the next article, by Ellis and Vasconcelos (2000). Anyone interested in tracing the history of facets and hypertext will, however, want to see the originals.

Type

a

0.0028047764 = product of:
  0.005609553 = sum of:
    0.005609553 = product of:
      0.011219106 = sum of:
        0.011219106 = weight(_text_:a in 5570) [ClassicSimilarity], result of:
          0.011219106 = score(doc=5570,freq=22.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.21126054 = fieldWeight in 5570, product of:
              4.690416 = tf(freq=22.0), with freq of:
                22.0 = termFreq=22.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0390625 = fieldNorm(doc=5570)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Dagobert Soergel opened the discussion by saying, "Users need our help to find and make sense of information." He said that he believes that classification can provide much of the support users need. He sees little distinction between the concept of a thesaurus and that of classification since both seek to provide structure for knowledge bases in order to facilitate information retrieval. Soergel's discussion of his concept of a multifunctional, multilingual thesaurus comprised the bulk of his presentation. This thesaurus would be a database of concepts, terms, and relationships which would include classification. In this context, classification has a much broader set of functions than it currently does in most American libraries where classification serves primarily as a method of shelf arrangement. The thesaurus Soergel envisions would lay out the semantic map of a field and could therefore be used as a learning tool or as a basis for research planning, or to assist users in clarifying terms and concepts. It could support indexing and searching and provide for the organization of knowledge for expert systems and other artificial intelligence applications. Among its other features, such a thesaurus could assist users in making sense of information by providing structured presentation of search results based on user needs and preferences, and it could enhance natural language processing capabilities such as automated indexing and abstracting and machine translation.

Type

a

0.0027127487 = product of:
  0.0054254974 = sum of:
    0.0054254974 = product of:
      0.010850995 = sum of:
        0.010850995 = weight(_text_:a in 97) [ClassicSimilarity], result of:
          0.010850995 = score(doc=97,freq=42.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20432885 = fieldWeight in 97, product of:
              6.4807405 = tf(freq=42.0), with freq of:
                42.0 = termFreq=42.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.02734375 = fieldNorm(doc=97)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Organizing and providing access to the resources an the Internet has been a problem area in spite of the availability of sophisticated search engines and other Software tools. There have been several attempts to organize the resources an the World Wide Web. Some of them have tried to use traditional library classification schemes such as the Library of Congress Classification, the Dewey Decimal Classification and others. However there is a need to assign proper subject headings to them and present them in a logical or hierarchical sequence to cater to the need for browsing. This paper attempts to describe an experimental system designed to organize and provide access to web documents using a faceted pre-coordinate indexing system based an the Deep Structure Indexing System (DSIS) derived from POPSI (Postulate based Permuted Subject Indexing) of Bhattacharyya, and the facet analysis and chain indexing system of Ranganathan. A prototype Software System has been designed to create a database of records specifying Web documents according to the Dublin Core and to input a faceted subject heading according to DSIS. Synonymous terms are added to the Standard terms in the heading using appropriate symbols. Once the data are entered along with a description and the URL of the web document, the record is stored in the System. More than one faceted subject heading can be assigned to a record depending an the content of the original document. The System stores the Surrogates and keeps the faceted subject headings separately after establishing a link. The search is carried out an index entries derived from the faceted subject heading using the chain indexing technique. If a single term is Input, the System searches for its presence in the faceted subject headings and displays the subject headings in a sorted sequence reflecting an organizing sequence. If the number of retrieved Keadings is too large (running into more than a page) the user has the option of entering another search term to be searched in combination. The System searches subject headings already retrieved and looks for those containing the second term. The retrieved faceted subject headings can be displayed and browsed. When the relevant subject heading is selected the system displays the records with their URLs. Using the URL, the original document an the web can be accessed. The prototype system developed in a Windows NT environment using ASP and a web server is under rigorous testing. The database and Index management routines need further development.
An interesting but somewhat confusing article telling how the writers described web pages with Dublin Core metadata, including a faceted classification, and built a system that lets users browse the collection through the facets. They seem to want to cover too much in a short article, and unnecessary space is given over to screen shots showing how Dublin Core metadata was entered. The screen shots of the resulting browsable system are, unfortunately, not as enlightening as one would hope, and there is no discussion of how the system was actually written or the technology behind it. Still, it could be worth reading as an example of such a system and how it is treated in journals.

Footnote

Vgl. auch: Devadason, F.J.: Facet analysis and Semantic Web: musings of a student of Ranganathan. Unter: http://www.geocities.com/devadason.geo/FASEMWEB.html#FacetedIndex.

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 107) [ClassicSimilarity], result of:
          0.0108246 = score(doc=107,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 107, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=107)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 108) [ClassicSimilarity], result of:
          0.0108246 = score(doc=108,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 108, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=108)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 109) [ClassicSimilarity], result of:
          0.0108246 = score(doc=109,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 109, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=109)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 113) [ClassicSimilarity], result of:
          0.0108246 = score(doc=113,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 113, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=113)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 156) [ClassicSimilarity], result of:
          0.0108246 = score(doc=156,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 156, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=156)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 1905) [ClassicSimilarity], result of:
          0.0108246 = score(doc=1905,freq=2.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 1905, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.125 = fieldNorm(doc=1905)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 2108) [ClassicSimilarity], result of:
          0.0108246 = score(doc=2108,freq=8.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 2108, product of:
              2.828427 = tf(freq=8.0), with freq of:
                8.0 = termFreq=8.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0625 = fieldNorm(doc=2108)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

It will not be an easy task to automate the Library of Congress Classification schedules because it is a very large system and also because it developed long before automation. The designers were creating a system for shelving books effiently and had not even imagined the constraints imposed by automation. A number of problems and possible solutions are discussed. The MARC format proposed for classification has some serious problems which are identified

Type

a

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 2417) [ClassicSimilarity], result of:
          0.0108246 = score(doc=2417,freq=8.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 2417, product of:
              2.828427 = tf(freq=8.0), with freq of:
                8.0 = termFreq=8.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0625 = fieldNorm(doc=2417)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

An online series of articles explaining controlled vocabularies and, in particular, faceted classification. It is not yet finished, but what they have covered is very well done, practical and informative, with useful advice and a full treatment. It is worth reading now, and when they actually get to performing facet analysis and making a faceted system, it will make a very useful reference.

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 2461) [ClassicSimilarity], result of:
          0.0108246 = score(doc=2461,freq=8.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 2461, product of:
              2.828427 = tf(freq=8.0), with freq of:
                8.0 = termFreq=8.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0625 = fieldNorm(doc=2461)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

An online series of articles explaining controlled vocabularies and, in particular, faceted classification. It is not yet finished, but what they have covered is very well done, practical and informative, with useful advice and a full treatment. It is worth reading now, and when they actually get to performing facet analysis and making a faceted system, it will make a very useful reference.

0.00270615 = product of:
  0.0054123 = sum of:
    0.0054123 = product of:
      0.0108246 = sum of:
        0.0108246 = weight(_text_:a in 2469) [ClassicSimilarity], result of:
          0.0108246 = score(doc=2469,freq=8.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20383182 = fieldWeight in 2469, product of:
              2.828427 = tf(freq=8.0), with freq of:
                8.0 = termFreq=8.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0625 = fieldNorm(doc=2469)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

An online series of articles explaining controlled vocabularies and, in particular, faceted classification. It is not yet finished, but what they have covered is very well done, practical and informative, with useful advice and a full treatment. It is worth reading now, and when they actually get to performing facet analysis and making a faceted system, it will make a very useful reference.

0.0026849252 = product of:
  0.0053698504 = sum of:
    0.0053698504 = product of:
      0.010739701 = sum of:
        0.010739701 = weight(_text_:a in 3991) [ClassicSimilarity], result of:
          0.010739701 = score(doc=3991,freq=14.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20223314 = fieldWeight in 3991, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046875 = fieldNorm(doc=3991)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Although researchers have theorized the critical importance of classification in the organization of information, the classification approach seems to have given way to the alphabetical subject approach in retrieval tools widely used in libraries, and research an how users utilize classification or classification-like arrangements in information seeking has been scant. To better understand whether searchers consider classificatory structures a viable alternative to information retrieval, this article reports an a study of how 24 library and information science students used Yahoo! directories, a popular search service resembling classification, in completing an assigned simple task. Several issues emerged from the students' reporting of their search process and a comparison between hierarchical navigation and keyword searching: citation order of facets, precision vs. recall, and other factors influencing searchers' successes and preferences. The latter included search expertise, knowledge of the discipline, and time required to complete the search. Without a definitive conclusion, we suggest a number of directoons for further research.

Type

a

0.0026849252 = product of:
  0.0053698504 = sum of:
    0.0053698504 = product of:
      0.010739701 = sum of:
        0.010739701 = weight(_text_:a in 534) [ClassicSimilarity], result of:
          0.010739701 = score(doc=534,freq=14.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20223314 = fieldWeight in 534, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046875 = fieldNorm(doc=534)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

In free classification, each concept is expressed by a constant notation, and classmarks are formed by free combinations of them, allowing the retrieval of records from a database by searching any of the component concepts. A refinement of free classification is freely faceted classification, where notation can include facets, expressing the kind of relations held between the concepts. The Integrative Level Classification project aims at testing free and freely faceted classification by applying them to small bibliographical samples in various domains. A sample, called the Dandelion Bibliography of Facet Analysis, is described here. Experience was gained using this system to classify 300 specialized papers dealing with facet analysis itself recorded on a MySQL database and building a Web interface exploiting freely faceted notation. The interface is written in PHP and uses string functions to process the queries and to yield relevant results selected and ordered according to the principles of integrative levels.

Type

a

0.0026742492 = product of:
  0.0053484985 = sum of:
    0.0053484985 = product of:
      0.010696997 = sum of:
        0.010696997 = weight(_text_:a in 2478) [ClassicSimilarity], result of:
          0.010696997 = score(doc=2478,freq=20.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20142901 = fieldWeight in 2478, product of:
              4.472136 = tf(freq=20.0), with freq of:
                20.0 = termFreq=20.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0390625 = fieldNorm(doc=2478)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

Three pre-web articles about using hypertext for knowledge representation. Duncan discusses how to use graphical, hypertext displays (she used Xerox PARC's NoteCards on a Xerox 1186 workstation) along with concept maps and facet analysis, a combination that would now be done with topic maps. The screen shots of her graphical displays are quite interesting. Her interest in facets is in how to use them to show things to different people in different ways, for example, so that experts can enter knowledge into a system in one way while novices can see it in another. Duncan found that facet labels (e.g. Process and Product) prompted the expert to think of related concepts when inputting data, and made navigation easier for users. Facets can be joined together, e.g. "Agents (causing) Process," leading to a "reasoning system." She is especially interested in how to show relstionships between two things: e.g., A causes B, A uses B, A occurs in B. This is an important question in facet theory, but probably not worth worrying about in a small online classification where the relations are fixed and obvious. These articles may be difficult to find, in which case the reader can find a nice sumary in the next article, by Ellis and Vasconcelos (2000). Anyone interested in tracing the history of facets and hypertext will, however, want to see the originals.

Type

a

0.0026742492 = product of:
  0.0053484985 = sum of:
    0.0053484985 = product of:
      0.010696997 = sum of:
        0.010696997 = weight(_text_:a in 4327) [ClassicSimilarity], result of:
          0.010696997 = score(doc=4327,freq=20.0), product of:
            0.053105544 = queryWeight, product of:
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.046056706 = queryNorm
            0.20142901 = fieldWeight in 4327, product of:
              4.472136 = tf(freq=20.0), with freq of:
                20.0 = termFreq=20.0
              1.153047 = idf(docFreq=37942, maxDocs=44218)
              0.0390625 = fieldNorm(doc=4327)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

The suggestion that classifications for retrieval should be constructed automatically raises some serious problems concerning the sorts of classification which are required, and the way in which formal classification theories should be exploited, given that a retrieval classification is required for a purpose. These difficulties have not been sufficiently considered, and the paper therefore attempts an analysis of them, though no solution of immediate application can be suggested. Starting with the illustrative proposition that a polythetic, multiple, unordered classification is required in automatic thesaurus construction, this is considered in the context of classification in general, where eight sorts of classification can be distinguished, each covering a range of class definitions and class-finding algorithms. The problem which follows is that since there is generally no natural or best classification of a set of objects as such, the evaluation of alternative classifications requires either formal criteria of goodness of fit, or, if a classification is required for a purpose, a precises statement of that purpose. In any case a substantive theory of classification is needed, which does not exist; and since sufficiently precise specifications of retrieval requirements are also lacking, the only currently available approach to automatic classification experiments for information retrieval is to do enough of them

Type

a