home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!decwrl!purdue!yuma!ld231782
- From: ld231782@LANCE.ColoState.Edu (L. Detweiler)
- Newsgroups: comp.ai
- Subject: online thesaurus + travelling salesman = word spectrum
- Message-ID: <Sep02.180231.79999@yuma.ACNS.ColoState.EDU>
- Date: 2 Sep 92 18:02:31 GMT
- Sender: news@yuma.ACNS.ColoState.EDU (News Account)
- Organization: Colorado State University, Fort Collins, CO 80523
- Lines: 41
-
-
- - - -
-
- Hello. Since the subject of online thesauruses pops up here and is in the
- FAQ, I thought I would ask a question of people involved in this area.
-
- Suppose I have a (large) set of words. A thesaurus can be represented as
- a graph with words as vertices and edges denoting synonymnity. Now, suppose
- the shortest path can be found that visits all vertices. In a sense this
- collapses the multidimensional graph into a linear listing. The
- structure of this listing would be very interesting from the point of view
- of a lexicographer. It would be a listing of all the words in the set
- in the order of `shades of meaning', and even gives a metric for this--
- the similarity of two words can be defined as their distance from each
- other in the list.
-
- In a sense, this could be used to give a new kind of reference in addition
- to a thesaurus or dictionary, combining some of the best features of both.
- If a person wants to learn the meaning of a word, s/he looks in the index
- to find the location of the word. Then, looking it up, examines neighboring
- words, hopefully more familiar, to grasp at the meaning of the newer word.
- It could be used as a thesaurus in a similar fashion.
-
- Depending on the scheme, weights may be assigned to edges of the graph as
- either unit lengths or some other measure based on the "nearness" of
- two words (for example, the average distance of all paths connecting them).
- If the weights of edges are nonuniform, the endpoints of the optimal path
- may be determined uniquely. It would be interesting to speculate on what two
- words would fall at the endpoints--`good'/`evil'? `true'/`false'?
- `love'/`hate'?
-
- The question, then--has anybody attempted to implement such a scheme? I
- know that Travelling Salesman is apropos and in NP, so that the optimum
- path is intractable for larger sets, but even approximations would be useful.
- The project may be feasible and even lucrative to an inventor, considering
- the importance and ubiquity of language reference materials.
-
-
- --
-
- ld231782@longs.LANCE.ColoState.EDU
-