Search (3 results, page 1 of 1)

0.0089465715 = product of:
  0.026839713 = sum of:
    0.026839713 = weight(_text_:search in 3634) [ClassicSimilarity], result of:
      0.026839713 = score(doc=3634,freq=2.0), product of:
        0.1747324 = queryWeight, product of:
          3.475677 = idf(docFreq=3718, maxDocs=44218)
          0.05027291 = queryNorm
        0.15360467 = fieldWeight in 3634, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.475677 = idf(docFreq=3718, maxDocs=44218)
          0.03125 = fieldNorm(doc=3634)
  0.33333334 = coord(1/3)

Abstract

In the continuing search for criteria for determining the form of compound headings (i.e., headings containing more than one word), many authors have attempted to deal with the problem of entry element and citation order. Among the proposed criteria are Cutter's concept of "significance," Kaiser's formula of "concrete/process," Prevost's "noun rule," and Farradane's categories of relationships*' (q.v.). One of the problems in applying the criteria has been the difficulty in determining what is "significant," particularly when two or more words in the heading all refer to concrete objects. In the following excerpt from Subject Catalogues: Headings and Structure, a widely cited book an the alphabetical subject catalog, E. J. Coates proposes the concept of "term significance," that is, "the word which evokes the clearest mental image," as the criterion for determining the entry element in a compound heading. Since a concrete object generally evokes a clearer mental image than an action or process, Coates' theory is in line with Kaiser's theory of "concrete/process" (q.v.) which Coates renamed "thing/action." For determining the citation order of component elements in a compound heading where the elements are equally "significant" (i.e., both or all evoking clear mental images), Coates proposes the use of "term relationship" as the determining factor. He has identified twenty different kinds of relationships among terms and set down the citation order for each. Another frequently encountered problem related to citation order is the determination of the entry element for a compound heading which contains a topic and a locality. Entering such headings uniformly under either the topic or the locality has proven to be infeasible in practice. Many headings of this type have the topic as the main heading, subdivided by the locality; others are entered under the locality as the main heading with the topic as the subdivision. No criteria or rules have been proposed that ensure consistency or predictability. In the following selection, Coates attempts to deal with this problem by ranking the "main areas of knowledge according to the extent to which they appear to be significantly conditioned by locality." The theory Coates expounded in his book was put into practice in compiling the British Technology Index for which Coates served as the editor from 1961 to 1977.

0.0089465715 = product of:
  0.026839713 = sum of:
    0.026839713 = weight(_text_:search in 3641) [ClassicSimilarity], result of:
      0.026839713 = score(doc=3641,freq=2.0), product of:
        0.1747324 = queryWeight, product of:
          3.475677 = idf(docFreq=3718, maxDocs=44218)
          0.05027291 = queryNorm
        0.15360467 = fieldWeight in 3641, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          3.475677 = idf(docFreq=3718, maxDocs=44218)
          0.03125 = fieldNorm(doc=3641)
  0.33333334 = coord(1/3)

Abstract

The second Cranfield experiment (Cranfield II) in the mid-1960s challenged assumptions held by librarians for nearly a century, namely, that the objective of providing subject access was to bring together all materials an a given topic and that the achieving of this objective required vocabulary control in the form of an index language. The results of Cranfield II were replicated by other retrieval experiments quick to follow its lead and increasing support was given to the opinion that natural language information systems could perform at least as effectively, and certainly more economically, than those employing index languages. When the results of empirical research dramatically counter conventional wisdom, an obvious course is to question the validity of the research and, in the case of retrieval experiments, this eventually happened. Retrieval experiments were criticized for their artificiality, their unrepresentative sampies, and their problematic definitions-particularly the definition of relevance. In the minds of some, at least, the relative merits of natural languages vs. indexing languages continued to be an unresolved issue. As with many eitherlor options, a seemingly safe course to follow is to opt for "both," and indeed there seems to be an increasing amount of counsel advising a combination of natural language and index language search capabilities. One strong voice offering such counsel is that of Robert Fugmann, a chemist by training, a theoretician by predilection, and, currently, a practicing information scientist at Hoechst AG, Frankfurt/Main. This selection from his writings sheds light an the capabilities and limitations of both kinds of indexing. Its special significance lies in the fact that its arguments are based not an empirical but an rational grounds. Fugmann's major argument starts from the observation that in natural language there are essentially two different kinds of concepts: 1) individual concepts, repre sented by names of individual things (e.g., the name of the town Augsburg), and 2) general concepts represented by names of classes of things (e.g., pesticides). Individual concepts can be represented in language simply and succinctly, often by a single string of alphanumeric characters; general concepts, an the other hand, can be expressed in a multiplicity of ways. The word pesticides refers to the concept of pesticides, but also referring to this concept are numerous circumlocutions, such as "Substance X was effective against pests." Because natural language is capable of infinite variety, we cannot predict a priori the manifold ways a general concept, like pesticides, will be represented by any given author. It is this lack of predictability that limits natural language retrieval and causes poor precision and recall. Thus, the essential and defining characteristic of an index language ls that it is a tool for representational predictability.

0.0039732503 = product of:
  0.01191975 = sum of:
    0.01191975 = product of:
      0.0238395 = sum of:
        0.0238395 = weight(_text_:22 in 3644) [ClassicSimilarity], result of:
          0.0238395 = score(doc=3644,freq=2.0), product of:
            0.17604718 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.05027291 = queryNorm
            0.1354154 = fieldWeight in 3644, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.02734375 = fieldNorm(doc=3644)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Footnote

Original in: Information retrieval today: papers presented at an Institute conducted by the Library School and the Center for Continuation Study, University of Minnesota, Sept. 19-22, 1962. Ed. by Wesley Simonton. Minneapolis, Minn.: The Center, 1963. S.21-36.