Comparing Thesaurus Table Variants

Here is a summary of the similarities and differences across the three types of thesaurus tables:

•

The sole purpose of a classic thesaurus table is to permit “substitutions”: the matching of lexically dissimilar terms between query and record.

•

Unlike the other two forms of thesauri, the classic thesaurus allows only a single penalty that applies to all classes within the thesaurus rather than a penalty or weight per class.

•

The primary purpose of a weighted dictionary is to specify the relative importance of certain terms. The term weight is a structural weight, not a penalty. That is, it does not necessarily lower the score of a record but simply adjusts the influence of a particular term on the score, whether the term is matched or unmatched. (Unmatched material in the query influences the value of the normal score, unmatched material in the record influences the reverse score, and unmatched material in either the query or the record influences the symmetric score.)

•

A weighted dictionary can define equivalences as a secondary feature (For example, equating abbreviated and non-abbreviated forms of a lightweight term). The weighted dictionary, however, provides no method for applying a penalty to these substitutions.

•

Since the primary purpose of the weighted dictionary is the semantic weighting of individual terms rather than substitution of equivalent terms, it is permissible for a weighted dictionary to contain equivalence classes containing only a single term (and the term weight).

•

The combined thesaurus table provides the ability to specify any combination of both types of relationships (substitutions with penalties, plus semantic weighting) in a single table structure. If an individual class in a combined thesaurus specifies a semantic term weight of 1.0, then it functions only to specify possible substitutions, as specified in the classic thesaurus table. If an individual class in a combined thesaurus table specifies a semantic term weight of other than 1.0, this weight adjusts the influence of the term whenever it occurs (matched or unmatched). If the same class contains more than one element, substitutions might occur, which are penalized according to the thesaurus weight for the class. Unlike the classic thesaurus, the combined thesaurus allows a different substitution penalty to be specified for each equivalence class.