Natural versus Indexing Language in Chemical Documentation

Abstract

In any literature search one must be able to predict which modes of expression have been used to represent the concepts of interest in the search file. It is these lingual expressions which will be looked up or will constitute the search parameters in a mechanized search. In the case of general concepts and statements, uncontrolled natural language lacks the required representational predictability, as is demonstrated, for example, for classes of chemical compounds and for types of reactions. Hence, considerable loss of relevant information is bound to occur in searches for these concepts in free text files, regardless of how advanced a computer program may be. By restricting the modes of expression as is achieved by means of an indexing language and its reliable employment, predicability is increased and, consequently, information loss is drastically reduced.-In contrast, the modes of expression for indiuidual concepts (e. g . individual chemical substances, authors, institutions) are, by their very nature, often sufficiently predictable in uncontrolled natural language. In these cases, the value of translating them into indexing-language expressions is often ques-Wissenschaftliche Dokumentation, Hoechst AG D-6230 Frankfurt am Main 80 (Germany) 608 0 Verlag Chernie GmbH. 6940 Weinheim, 1982 0570-0833/82/0808-0608 $ 02.50/0 Angew. Chem. Int. Ed. Engl. 21 (1982) 608-616