[Corpora-List] Summary: Specificty and Similarity of Words

Nuno Seco nseco at xldb.di.fc.ul.pt
Fri Mar 18 12:39:00 CET 2005


I have recently found this paper:
http://crpit.com/confpapers/CRPITV38Yang.pdf

I've only skimmed through it, but it also seems relevant.

--
Nuno

On Thu, 2005-03-17 at 19:17 +0200, Ergin Altintas wrote:

> Dear corpra members;

>

> Sorry for my sending of the same message describing my questions about

> Specificty and Similarity of Words three times (on Wed, 16 Mar 2005

> 19:56:58 +0200). It was because of an error report returned by the

> maillist server which I couldn't understand because it wasn't in

> English.

>

> Any way there were 3 answers for my questions:

>

> 1. Nuno Seco referred to a work of his own. And this is what I was

> exactly looking for. I had a review of the paper, good work.

> http://afflatus.ucd.ie/~ebbis/seco.pdf

>

> 2. Mark Sanderson noted IDF and Karen Sparck Jones's ideas about

> specificty. They were mainly dependent on large copora. And that was

> not what I was looking for (yes I had not mentioned about "no

> corpora"). I wasnt famialiar to IDF and for the ones who consider to

> learn more about IDF's I think I should sahre this brief page having

> links to original resources is located here:

> http://www.soi.city.ac.uk/~ser/idf.html

>

> 3. Ronald P. Reck said, he used Ted Pedersen's

> WordNet::Similarity package sucessfully. Yes. I also use this

> package and its great to have such open source projects for a good

> starts in CL. (But in fact this wasn't an answer for my questions)

> http://www.d.umn.edu/~tpederse/similarity.html

>

> I am still looking for similar work as in (1).

>

> And one of my questions is not aswered yet exactly:

>

> How should similarity measures be evaluated? Is it enough to show the

> correlation to human judgements or shall we give practical

> results of application areas? (like WSD, IR, etc...)

>

> Thanks to everyone who had interest in my questions.

>

> Best Wishes,

>

>

> Ergin ALTINTAS

> Turkish Naval Academy






More information about the Corpora-archive mailing list