De Invloed Van Tekststructuur En Lexicale-Woordfrequentie Op Behoud Van Informatie In Referaten
Η.Α.J.M. Lamers |
Sectie Wetenschappelijke-informatieverzorging Landbouwuniversiteit Wageningen
Scientific information vanishes in the transfer process. Input losses point out as a success model: English articles published in American journals with keywords in the titles, with an English author abstract, tables, graphs, and a bibliography. Output losses demonstrate as a succes model: controlled descriptors properly handled from a thesaurus. These output losses require language knowledge, especially of English. Information was linguistically defined as lexical unit types. A well-structured and an ill-structured text were each presented to 25 subjects to be abstracted. Each original text had 864 tokens and 178 lexical unit types. The abstracts from the well-structured text retained 28% of the lexical unit types; those from the ill-structured text 17%. The former kind contained 40% more lexical unit types than the latter kind. The former kind contained every lexical unit type with a frequency of 6 and more; the latter kind those with with a frequency of 9 and more. The lexical unit types in both kinds of abstracts correlated to those in the original text 0.98. Correlation of the lexical unit types in the abstracts to each other was 0.995. Hypotheses: a structure is a peremptory requirement for retention of information; the frequency needed is dependent on the structure; standardized keywords make retention easier.
Article language: Dutch