Table of Contents

NAME

wngroups - discussion of WordNet search code to group similar senses

DESCRIPTION

Standard dictionaries commonly group related senses of a word with expressions such as "the act, process, or result of X", or else they enter the act, process, and result senses as subheads under a single sense number. Unlike standard dictionaries, the default display for the WordNet browser is to show senses in order of frequency of use in the semantically tagged texts described in semcor(7WN) . The RELATIVES search displays similar senses of a word together. At the present time, nouns and some verbs are grouped.

Noun Groups

Three relations are used to group noun senses: cousins, sisters, and twins.

The cousin groupings exploit the hyponym relation in WordNet. Many WordNet nodes whose hyponyms bear a specific relation to each other have been identified. For example, the noun crab refers to an animal, as well as the edible meat of the animal. The same relation, that of "an animal and its edible meat", holds for lobster , chicken and most other matching strings under the food and animal nodes.

Another class of related senses in the noun hierarchy is called sisters. Sisters are matching strings that are both the immediate hyponyms of the same superordinate. For example, the noun flounder can refer to several kinds of flatfish .

The third grouping relation is called twins. These are synsets that have at least three words in common. For example, one sense of duo is a musical group and another is a musical composition. Both synsets contain duet , duette , and duo .

Transitivity is used to combine groups of overlapping senses into the largest sense groups possible.

Verb Groups

Some similar senses of verbs have been grouped by the lexicographers. This grouping is done statically in the lexicographer source files using the semantic pointer_symbol $ . As with the noun senses, transitivity is used to combine groups of overlapping senses into the largest sense groups possible.

Exception List

There are, of course, exceptions to all of the relations. For example, the noun coral is in both the animal and food hierarchies, but does not bear the relation of "an animal and its edible meat". Candidates for cousin and twin groupings are checked by hand and those that should not be grouped together are listed in the exception list file.

NOTES

Coverage of noun cousins is incomplete.

Coverage of verb groups is incomplete.

Groups of noun senses are determined at run-time - not statically - when the RELATIVES search is requested. Depending on the computer platform and the number of senses involved, this search may run slowly.

ENVIRONMENT VARIABLES

WNHOME
Base directory for WordNet. Unix default is /usr/local/wordnet1.6 , PC default is C:\wn16 , Macintosh default is : .
WNSEARCHDIR
Directory in which the WordNet database has been installed. Unix default is WNHOME/dict , PC default is WNHOME\dict , Macintosh default is :Database .

FILES

All files are in directory WNSEARCHDIR :
cousin.tops
pairs of noun top nodes (Unix and Macintosh)
cousin.tps
pairs of noun top nodes (PC)
cousin.exc
senses that should not be grouped

SEE ALSO

wn(1WN) , wnb(1WN) , wnsearch(3WN) , wndb(5WN) semcor(7WN) .


Table of Contents