AW: techniques for better as well as for good [was: Re: programaticallylocated.doc]

Interesting discussion, and good to bring the people together from different
perspectives and WGs.

I think there are many ways we can do this technically.  One of the major
reasons to prefer one to the other is its likeliness to be actually used in
small and large-scale Web projects.  It's the business case that matters.

(1) This project will only fly if it achieves major benefit for the business
of translating websites into other natural languages.  The translation
business is a growing business, and companies will want to save money by
automation.  Dictionaries that help to disambiguate meaning can greatly help
in this.  However, this makes it necessary to address not only word meaning,
but also sentence meaning (as Jason has already pointed out).

(2) It will be necessary to base dictionaries and their entries on commonly
accepted (semantic) concepts, so that translation engines can base their
translations on them.  One example is WordNet, which is used by the Concept
Coding Framework.

(3) It may not be (economically) feasible to require that an author make
sure that every word in their document is covered by a definition in the
cascaded dictionary list.  As a starting point, an author (or any 3rd party
providing support for translation or for certain user groups) may provide or
link definitions that are most needed for translating the content (and thus
are the most ambiguous ones).  


> -----Ursprüngliche Nachricht-----
> Von:
> [] Im Auftrag von Gregg Vanderheiden
> Gesendet: Dienstag, 10. August 2004 04:29
> An:; 'wai-xtech'
> Betreff: RE: techniques for better as well as for good [was: 
> Re: programaticallylocated.doc]
> This is good. But much more complicated.  I am not  sure that
> every website or author could figure out how to semantically 
> characterize their content. 
> I do think that we need to not restrict connection to other
> dictionaries. But that seems to be a user agent issue.  
> RE this categorization.....   interesting.   I wonder how 
> easy we can make
> it- and how effective.
> Gregg
>  -- ------------------------------
> Gregg C Vanderheiden Ph.D. 
> Professor - Ind. Engr. & BioMed Engr.
> Director - Trace R & D Center 
> University of Wisconsin-Madison 
> -----Original Message-----
> From:
> [] On Behalf Of Jason White
> Sent: Monday, August 09, 2004 7:27 PM
> To: wai-xtech
> Subject: RE: techniques for better as well as for good [was: 
> Re: programaticallylocated.doc]
> Responding to Gregg's original proposal and Al's comments,
> often the author-specified dictionaries may not be what the 
> user needs. Rather, the user typically requires a certain 
> type of dictionary, for example a multilingual dictionary 
> providing translations or definitions in a preferred 
> language. These can be radically different from whatever 
> dictionaries the author has chosen. Furthermore, as Al 
> indicated, a list of available definitions is insufficient; 
> what is required, rather, is the correct definition.
> My proposal is to attach metadata to the content, or parts of
> the content, that specifies subject descriptors taken from a 
> controlled vocabulary, such as appear in bibliographic 
> records. Downstream software (i.e., server-side or 
> client-side tools) can then make use of this classification 
> scheme, together with the user's preferences regarding 
> languages and dictionaries, to establish a suitable list of 
> dictionaries to be looked up, and an appropriate search 
> order. A typical dictionary also categorizes definitions by 
> subject matter. To the extent that this is achieved, it 
> enables the correct definition to be programmatically  determined.
> For example, the word "field" has an entirely different
> meaning in an article on agriculture to that which it conveys 
> in an algebra textbook. If the content is labelled in 
> metadata as related to algebra, the retrieval software can 
> either search a mathematics dictionary for the precise, 
> technical definition, or find it in a more general 
> dictionary, provided that the latter is appropriately marked 
> up to distinguish mathematical definitions from the remaining 
> alternatives.
> This suggested approach could of course be combined with
> direct specification of dictionaries by the author; the two 
> proposals are by no means mutually exclusive.

Received on Tuesday, 10 August 2004 08:55:57 UTC