[Corpora-List] Network analysis

[jgorman2 at cs.usyd.edu.au]

Hi Bob,

I played around with small-world graphs and thesauri. I wrote a small c++
tool to get graph statistics and generate random graphs for comparative
analysis. You're probably better off with one of the earlier suggested
tools (Pajek or DIMAP) since they are considerably more mature, but if you
need something you can run at the command line and hack about in, it's
available and I can probably spare some time to help if needed.

Since there's no website for it I'll have to point you to the paper that
uses it:
The Topology of Synonymy and Homonymy Networks, Cognitive Aspects of
Computational Language Acquisition, ACL 07
http://www.aclweb.org/anthology-new/W/W07/W07-0610.pdf

James


> Prof. Wilks,
> I remember seeing an early version of Plate's analysis in
> "Computational Lexicography for NLP" (Boguraev & Briscoe). I'm
> interested primarily in the notion that the semantic "hubs" in a
> network analysis would turn out to be central concepts in leveraging
> cognitive development. For example, "cause" (influence, power, etc.)
> or "instrument" (tool, implement, etc.) would surely be central hubs
> in our semantic domains. The interesting question for me is how two
> strategies of differentiation occur: (1) enriching a term with
> multiple senses; and (2) differentiating among multiple terms in a
> field of similar terms.  It has been proposed that education can be
> enriched by in depth study of a core set of terms (Gardner, The
> Disciplined Mind). And it has also been proposed that its important
> to have a wide ranging vocabulary for concept discrimination. Both
> are probably true. But educational specialists have no systematic
> criteria for choosing the keywords in education.
>
> Thanks for your response. Any advice for this novice is appreciated.
> Bob

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora