Syntactical devices
Syntactical devices are tools aiming at the increase of precision in information retrieval when the lack of precision is caused by grammatical issues.
Example: Searching literature about
"children of blind parents" may lack precision if searched by
Boolean intersection by combining the
search terms
"children" AND "blind" AND "parents" because most literature is about
parents of blind children.
Based on such experiences work has been
done to introduce syntactical devices into
information retrieval languages.
Early examples were "Roles" and "links", later
PRECIS-strings.
"Roles" (also termed "relational indicators") are the assignment of an indicator to an indexing term, which indicates the specific kind of relation which exists between the indexing terms.
"Links" is a technique in which index terms may be assigned to
different fields or sub-fields according to their group or class in order to
limit a search to those fields.
Both "roles" and "links" were soon given up again. Spang-Hanssen (1976, p. 22) explains why:
"Thus the idea of a syntactic pattern is not particularly fit for indexing. If an indexer concentrates on a particular syntactic pattern and, accordingly, ascribes particular syntactic roles to the terms in question, the indexing may turn out to be too narrow, to put too many restrictions on the combinability of terms, "to introduce too much structure". The result will be unsatisfactory retrieval, since a document may - for reasons of role structure - not be retrieved, in spite of the fact that it might be relevant from a less restrictive point of view". (Spang-Hanssen, 1976, p. 22).
In spite of those early experiences, the value of syntactical devices may still
be considered an open issue in IR-research. The quick given-up may partly be
explained by the costs of applying such devises and partly by the rapid
development of competing technologies.
Green (1995) writes: "To be fully effective, the expression of conceptual syntagmatic relationships must comply with criteria of systematicity, complexity, efficiency and naturalness. Unfortunately, the complex interaction of natural language means of expressing these relationships (lexicalisation, word order, function words and morphosyntactic cases) causes them to fail the systematicity criterion. Most document retrieval system means of expressing conceptual syntagmatic relationships (as exemplified by various term co-occurrence techniques, links and role indicators) fail to comply with this and other of the criteria. Only gestalt structures simultaneously representing relationships, participants and roles (for example, frames) conform fully to the criterial checklist. "
Literature:
Austin, D. (1984). PRECIS: a Manual of Concept Analysis and Subject
Indexing. 2. ed. London: British Library.
Austin, D. & Digger, J. A. (1977). PRECIS: the Preserved Context Index System. Library Resources and Technical Services, 21(1), 13-30.
Green, R. (1995). The expression of
conceptual syntagmatic relationships: A
comparative survey. Journal of Documentation, 51(4), 315-338.
Spang-Hanssen, H. (1976). Roles and links compared with grammatical relations in natural language. København: DTL.
See also: Natural Language Processing; Paradigmatic-syntagmatic (Lifeboat for KO); PRECIS (Lifeboat for KO).
Birger Hjørland
Last edited: 25-03-2006