FOA Home | UP: Inference beyond the \Index


Keyword structures

Most of what we have said about the Index relation (e.g., as part of the vector space model) assumes that keywords are simply a set of features. But beyond simply providing access to the retrieval of documents, the fact that keywords are \rikmeaning-ful objects in their own right means that we can analyze relationships among keywords directly.

THESAURI are structured representations of relations among keywords. Common relations represented in thesauri include:

\item \textbf{Broader term/narrower term (BT/NT)}; these capture hierarchic relations, generally between a kind of semantics, sometimes part whole relation. \item \textbf{Related term (RT)} capturing synonym or quasi-synonym relationships. \item \textbf{Use for (UF)} capturing a preferred, conventional or authoritative term over possible alternatives.

One of the most extensive examples of such a representation is the MESH (MEdical Subject Headings) thesaurus, part of the National Library of Medicine's extensive PubMed system. FIgure (figure) shows the term {\tt LYMPHOMA} within the MESH thesaurus. expanded in the second two tree locations.} example Hierarchic BT/NT relations are shown as indentation. Because this thesaurus allows a single keyword to fit in multiple places in the hierarchy (e.g., treating {\tt LYMPHOMA} as a kind of {\tt NEOPLASM} as well as a kind of {\tt IMMUNOLOGIC DISEASE} ) this browser shows the term as part of three separate paths; note that the children (narrower terms) of {\tt LYMPHOMA} are repeated at each location.

Subsections


Top of Page | UP: Inference beyond the \Index | ,FOA Home


FOA © R. K. Belew - 00-09-21