I still think that a homonym(s) property is relevant here, but maybe only <a href="http://en.wikipedia.org/wiki/Homograph">homographs</a> (AKA homoglyphs) for now. Homograph(s) (words with the same spelling but different meanings) should be really easy to populate after the initial load (e.g., IF synset AND name = name THEN homograph).This process might also help identify some merge candidates.<br>
<br>-Ed<br><br><br>On Tue, Sep 1, 2009 at 3:26 PM, Jeff Prucher <<a href="mailto:jeff@metaweb.com">jeff@metaweb.com</a>> wrote:<br>><br>> OK, let's see if we can agree on what aspects of synsets to model initially,<br>
> so we can go ahead with a load. (Jamie tells me that he was experimenting<br>> with WordNet awhile back and already has some loading scripts.)<br>><br>> I think Iain's Synonym Set is a good place to start:<br>
> <<a href="http://www.freebase.com/type/schema/base/writing/symset">http://www.freebase.com/type/schema/base/writing/symset</a>><br>><br>> I think the existing properties for hypernym and hyponym, should definitely<br>
> be included. I'd also vote for a synonym property. WordNet doesn't seem to<br>> have antonyms, so I'm not sure whether we want to keep that property for now<br>> or not (if we do, the expected type should be synset). As long as we're<br>
> doing this, holonyms and meronyms would presumably be simple as well<br>> (holonym="is a part of", meronym="has parts"), although probably less<br>> important.<br>><br>> Lexical category is probably necessary for now, but I think we should hold<br>
> off on Morphology.<br>><br>> Any other thoughts?<br>><br>> Jeff<br>><br>> > -----Original Message-----<br>> > From: <a href="mailto:data-modeling-bounces@freebase.com">data-modeling-bounces@freebase.com</a><br>
> > [mailto:<a href="mailto:data-modeling-bounces@freebase.com">data-modeling-bounces@freebase.com</a>] On Behalf Of Jeff Prucher<br>> > Sent: Thursday, August 27, 2009 11:45 AM<br>> > To: 'Freebase data modeling mailing list'<br>
> > Subject: Re: [Data-modeling] English Words<br>> ><br>> ><br>> ><br>> > > -----Original Message-----<br>> > > From: <a href="mailto:data-modeling-bounces@freebase.com">data-modeling-bounces@freebase.com</a><br>
> > > [mailto:<a href="mailto:data-modeling-bounces@freebase.com">data-modeling-bounces@freebase.com</a>] On Behalf Of Scott Meyer<br>> ><br>> > <Lots of interesting stuff omitted here about which I will<br>
> > comment in more depth when I have more time to do so.><br>> ><br>> > ><br>> > > Anyway... If you buy all this, then it seems reasonable to load<br>> > > Wordnet<br>> > > (synsets) first, then add pronunciation later.<br>
> ><br>> > I agree -- let's do the synsets first (skipping the issue of<br>> > symbols entirely for now). (Not that we shouldn't continue<br>> > the other discussions -- we can do it in parallel!)<br>
> ><br>> > Jeff<br>> ><br>> > _______________________________________________<br>> > Data-modeling mailing list<br>> > <a href="mailto:Data-modeling@freebase.com">Data-modeling@freebase.com</a><br>
> > <a href="http://lists.freebase.com/mailman/listinfo/data-modeling">http://lists.freebase.com/mailman/listinfo/data-modeling</a><br>> ><br>><br>> _______________________________________________<br>> Data-modeling mailing list<br>
> <a href="mailto:Data-modeling@freebase.com">Data-modeling@freebase.com</a><br>> <a href="http://lists.freebase.com/mailman/listinfo/data-modeling">http://lists.freebase.com/mailman/listinfo/data-modeling</a><br><br>