[Data-modeling] Products with ingredients

Jeff Prucher jeff at metaweb.com
Wed Jun 24 21:17:54 UTC 2009


There are three examples upthread that led to the phylogeny pattern, each of
which is a slightly different case:

(variety) <--> (generalization)
Milled corn <--> Corn
Sodium lauryl sulfate (from coconut oil) <--> Sodium lauryl sulfate Enriched
flour (foo, bar, bazz, fazz) <--> Enriched flour

Faye's division fits this pretty well: 
Milled corn is derived from corn; SLS (from coconut) is a variety of SLS,
and is also derived from coconut; enriched flour (etc., usw) is a variety of
enriched flour. (Reviewing this thread, I note that Ed suggested a Processed
Ingredient type way back at the outset.)  

The big question is, would we be asking for trouble by adding a parent/child
relationship to this, in addition to the two phylogeny patterns?  Or should
we just punt it for now?

Jeff


> -----Original Message-----
> From: data-modeling-bounces at freebase.com
> [mailto:data-modeling-bounces at freebase.com] On Behalf Of Tom Morris
> Sent: Wednesday, June 24, 2009 10:37 AM
> To: Freebase data modeling mailing list
> Subject: Re: [Data-modeling] Products with ingredients
> 
> I'm with Faye.  It seems very weird to have rice flour and rice so 
> strongly related.  I don't consider rice flour to be a generalization 
> of rice at all.  About the only places where they would potentially 
> interchangeable would be for nutritional information or for allergies.  
> You might be able to substitute basmati rice for jasmine rice if you 
> didn't care too much about the difference in texture or maintaining 
> cultural authenticity, but if you substituted rice flour (of any 
> variety), you'd be in a whole heap of trouble.
>
> 
> The examples in the schema descriptions (yay for 
> descriptions!) seem to have the same problem.  You can get 
> lavender oil out of a lavender plant, but they aren't 
> generalizations of each other.  If anything, the 
> generalization would be aromatic oil or fragrance or something.
> 
> For most applications, it's more useful to have things linked 
> together because of common properties rather than because 
> they are made from the same source material or by the same process.
> 
> -1 for making this even more obscure by linking in Material.
> 
> Tom
> 
> On Tue, Jun 23, 2009 at 9:02 PM, Faye Harris <faye at metaweb.com> wrote:
> 
> 
> 	Very cool!
> 	
> 	That rice flour is called a "variety of" rice in the 
> schema is indeed very odd.
> 	
> 	Based on the sandbox examples, this schema seems to use 
> "variety of" for two types of relationships:
> 	    1) variety of, e.g. brown rice is a "variety of" rice
> 	    2) derived from, e.g. rice flour is "derived from" rice
> 	
> 	The former relationship is categorical, the latter 
> relates to post-processing.
> 	
> 	-- Faye 
> 
> 
> 
> 	Jeff Prucher wrote: 
> 
> 		OK, I've got the double-phylogeny pattern 
> working now. Take a look here:
> 		
> http://www.sandbox-freebase.com/type/schema/business/product_i
ngredient
> 		
> 		And here's a table view of the ingredients of a 
> breakfast cereal I found in
> 		the office kitchen:
> 		
> http://www.sandbox-freebase.com/view/user/jeff/default_domain/
views/cranberr
> 		y_almond_crunch_ingredients
> 		
> 		I'm not really happy with the "variety of" and 
> "generalization of" names,
> 		but I'm not coming up with anything better. Any 
> suggestions would be most
> 		welcome.
> 		
> 		Jeff
> 		
> 		  
> 
> 			-----Original Message-----
> 			From: data-modeling-bounces at freebase.com 
> 			
> [mailto:data-modeling-bounces at freebase.com] On Behalf Of Jeff Prucher
> 			Sent: Friday, June 19, 2009 11:42 AM
> 			To: Freebase data modeling mailing list
> 			Subject: Re: [Data-modeling] Products 
> with ingredients
> 			
> 			
> 			----- "Faye Harris" <faye at metaweb.com> 
> <mailto:faye at metaweb.com>  wrote:
> 			
> 			    
> 
> 				From: "Faye Harris" 
> <faye at metaweb.com> <mailto:faye at metaweb.com> 
> 				To: "Freebase data modeling 
> mailing list" 
> 				      
> 
> 			<data-modeling at freebase.com> 
> <mailto:data-modeling at freebase.com> 
> 			    
> 
> 				Sent: Thursday, June 18, 2009 
> 2:59:19 PM GMT -08:00 
> 				      
> 
> 			US/Canada Pacific
> 			    
> 
> 				Subject: Re: [Data-modeling] 
> Products with ingredients
> 				
> 				Jeff Prucher wrote:
> 				      
> 
> 					----- "Robert Cook" 
> <robert at metaweb.com> <mailto:robert at metaweb.com>  wrote:
> 					
> 					  
> 					        
> 
> 					One solution would be 
> to create a topic with a long name -- enter
> 					          
> 
> 				it
> 				      
> 
> 					exactly as it appears 
> on the label such as "Enriched flour -
> 					          
> 
> 				(wheat,
> 				      
> 
> 					niacin, iron, baby 
> powder, sawdust, DDT)".
> 					    
> 					          
> 
> 					This would answer. 
> Anyone else have any comments or 
> 					        
> 
> 			thoughts on this
> 			    
> 
> 				before I load the schema?
> 				      
> 
> 					  
> 					        
> 
> 				The main problem with this is 
> you can't arrive at the products that 
> 				use enriched flour by clicking 
> on a property link from a single 
> 				"enriched
> 				
> 				flour" topic. Rather, you have 
> to do a keyword search for products 
> 				based on matching all the 
> various "enriched flour - (foo, bar, bat, 
> 				baz)"
> 				ingredient topics with the 
> words "enriched" and "flour". 
> 				      
> 
> 			That's quite 
> 			    
> 
> 				a loss in queriability.> > 
> 				The schema is fine to get us 
> started, but we're still going 
> 				      
> 
> 			to try to
> 			    
> 
> 				put together some phylogeny 
> pattern in place (in the near future) 
> 				right?
> 				      
> 
> 		  
> 
> 			I plan to add a phylogeny pattern 
> before moving the schema to 
> 			freebase.com, which should help 
> queryability. It doesn't 
> 			address the fact that topics named 
> things like "enriched 
> 			flour (that, that, the other thing)" 
> are exceedingly ugly, 
> 			however (no-one said it was called 
> "prettybase.com", though). 
> 			 I was going to post a revised schema 
> to sandbox, with the 
> 			double-phylogeny pattern suggested by 
> Robert, but it got 
> 			horribly munged in the process. I'll 
> try to fix it, but it 
> 			might not be till next week.
> 			
> 			Jeff
> 			_______________________________________________
> 			Data-modeling mailing list
> 			Data-modeling at freebase.com
> 			
> http://lists.freebase.com/mailman/listinfo/data-modeling
> 			
> 			    
> 
> 		_______________________________________________
> 		Data-modeling mailing list
> 		Data-modeling at freebase.com
> 		http://lists.freebase.com/mailman/listinfo/data-modeling
> 		
> 		  
> 
> 
> 
> 	_______________________________________________
> 	Data-modeling mailing list
> 	Data-modeling at freebase.com
> 	http://lists.freebase.com/mailman/listinfo/data-modeling
> 	
> 	
> 
> 
> 



More information about the Data-modeling mailing list