SYNTAGMATIC RELATIONSHIPS IN INDEX LANGUAGES - A REASSESSMENT

Effective use of syntagmatic relationships in index languages has suff ered from inaccurate or incomplete characterization in both linguistic s and information science. A number of ''myths'' about syntagmatic rel ationships are debunked: the exclusivity of paradigmatic and syntagmat ic relationships, linearity as a defining characteristic of syntagmati c relationships, the restriction of syntagmatic relationships to surfa ce linguistic units, the limitation of syntagmatic relationship benefi ts in document retrieval to precision, and the general irrelevance of syntagmatic relationships for document retrieval. None of the mechanis ms currently used with index languages is powerful enough to achieve t he levels of precision and recall that the expression of conceptual sy ntagmatic relationships is in theory capable of. New designs for expre ssing these relationships in index languages will need to take into ac count such characteristics as their semantic nature, systematicity, ge neralizability, and constituent nature.