[Corpora-List] Fwd: RE: Distance & word context.

[J Washtell]

Costas,

Thank you very much for clarifying your interpretation. I will
certainly dig deeper into the relationships between these paradigms.

Cheers,

Justin Washtell
University of Leeds


Quoting Costas Gabrielatos <c.gabrielatos at lancaster.ac.uk>:

> Justin,
>
> The span will be variable in two ways:
> -- The size of the span will differ according to document length.
> -- The right-side and left-side spans will not be 'mirrored' (to coin a
> term); i.e. unless the node is exactly in the middle of the document
> word-wise, then the right-side and left-side span will not be the same -
> e.g. in a 1,000-word text you may have a span of [-347, +652] - if your node
> is word #348.
>
> Costas
>
>
> -----Original Message-----
> From: J Washtell [mailto:lec3jrw at leeds.ac.uk]
> Sent: 03 May 2008 13:33
> To: Costas Gabrielatos
> Subject: Re: [Corpora-List] Distance & word context.
>
> Thanks Costas!
>
> Do you mean variable, insomuch as the length of each document differs?
> i.e. when framed as a "window"-based analysis, the size of the span is not
> absolute?
>
> Of course there are many such cases of document-based frequency analyses,
> but that's an interesting comparison which I had not thought of in such
> terms.
>
> Justin Washtell
> University of Leeds
>
>
> Quoting Costas Gabrielatos <c.gabrielatos at lancaster.ac.uk>:
>
>> Perhaps this paper might be of help:
>>
>> Kim, M-C. & Choi, K-S. (1999). A comparison of collocation-based
>> similarity measures in query expansion. Information Processing &
>> Management 35(1), 19-30.
>>
>> Their focus is query expansion, and they consider collocates of
>> candidate query terms setting the whole text as the span. The
>> interesting twist, as I see it, is that the span is at the same time
> well-defined and variable.
>>
>> Costas Gabrielatos
>> Lancaster University
>>
>>
>> -----Original Message-----
>> From: corpora-bounces at uib.no [mailto:corpora-bounces at uib.no] On Behalf
>> Of J Washtell
>> Sent: 30 April 2008 22:50
>> To: corpora at uib.no
>> Subject: [Corpora-List] Distance & word context.
>>
>> Hello all,
>>
>> This list is stimulating as always. I feel it is my turn to throw some
>> questions around :-)
>>
>> Can anybody point me towards works (however old or new) that exploit
>> the distance between terms in a corpus (such as, but not restricted
>> to, the use of "distance-weighted" context windows). The specific
>> applications are not important; I am interested in any works that deal
>> with the concept of distance as opposed to (or in addition to) say
>> frequency counts or roles/positions within grammatical constructs.
>>
>> Related to this, I am also interested in any work that courts the
>> notion of link-distance between words and texts within hypertext
>> structures (such as the Web); again, specific applications are
> unimportant.
>>
>> Finally (for the moment :-)), does anybody know of any (perhaps more
>> linguistically oriented) works that discuss the existence/importance
>> of
>> *very* long range dependencies and associations in text (e.g.
>> Dear... Yours, Results... Conclusion, etc), and the role these play
>> when considering word context.
>>
>> Kindest regards to all,
>>
>> Justin Washtell
>> University of Leeds
>>
>> ----------------------------------------------------------------
>> This message was sent using IMP, the Internet Messaging Program.
>>
>>
>>
>> _______________________________________________
>> Corpora mailing list
>> Corpora at uib.no
>> http://mailman.uib.no/listinfo/corpora
>>
>
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
>
>
>



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



----- End forwarded message -----


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


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