Análisis de los descriptores de diferentes áreas del conocimiento indizadas en bases de datos del CSIC. Aplicación a la indización automática

The value of scientific articles titles and abstracts as sources of terms for document indexing is studied in relation with six knowledge areas: Library and Information Science, Medicine, Chemistry, Biology, Psychology and Physics, indexed in the databases ISOC, IME and ICYT of the CSIC. The sintagmatic structures of the indexing terms found in the field Descriptors is also examined, as well as the relation between the length of the documents and the number of descriptors. In order to do this, six searches were made in the databases for the six knowledge areas, and 450 bibliographical references were selected (75 for knowledge area), obtaining 2.077 descriptors; of these, 38,1% appear in the titles, in the abstracts or in both. With respect to the syntactic structures it was found that 41,9% were nouns , 32,3% are noun+adjective groups, and 11,8% are noun+noun groups, with a 14% for other different structures. Lastly, regarding the relationship between length of documents and number of descriptors, all possible combinations were found: short articles with a few descriptors, long articles with a small amount of descriptors, short articles with a important quantity of descriptors, and documents with a high number both of pages and descriptors The following conclusions can be raised from the data obtained: first, if the abstracts are not well made and the titles are not precise, they are not definitives sources for the extraction of concepts; second, the most common syntactic structures is the noun phrase , followed by noun+adjective and noun-noun : third, no significant relation is found between length of documents and number of descriptors assigned to it. Se estudia el valor de los títulos y resúmenes de los artículos científicos como fuentes suministradoras de términos para la indización de los documentos en seis áreas del conocimiento indizadas en las bases de datos ISOC, IME e ICYT del CSIC. Asimismo, se examina la estructura sintagmática de los términos de indización hallados en el campo Descriptores , y la posible relación entre el número de descriptores de un documento con la cantidad de páginas del mismo. Para tales fines se seleccionaron las áreas del conocimiento de Biblioteconomía y Documentación, Medicina, Química , Biología, Psicología y Física, y se realizaron seis búsquedas en estas bases de datos de las que seleccionamos 450 referencias bibliográficas (75 por área) proporcionando un total de 2.077 descriptores. El 38,1% de los descriptores asignados a dichos registros aparece en el título, resumen o en el título y resumen a la v