You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@accumulo.apache.org by Benson Margulies <bi...@gmail.com> on 2012/02/02 00:15:33 UTC

Modeling provenance or other 'facts about facts'

Reading "ENTITY-ATTRIBUTE AND GRAPH TABLES" in the Table Design page,
I'm musing about where to put information about information.

If 'A manager_of B' (with confidence 0.5, supported by documents a, b, c)

...

how would you recommend adding that to the scheme described there?

Re: Modeling provenance or other 'facts about facts'

Posted by Adam Fuchs <sc...@gmail.com>.
I think one of the relevant questions here is how do you want to use the
confidence and provenance information? If you want to filter on it then it
probably fits well in the value. If you want to select on it then you might
want to model it as additional edges in your graph table (watch out for hot
spots of you go this route). Doing both might also be an option.

Adam
On Feb 1, 2012 6:16 PM, "Benson Margulies" <bi...@gmail.com> wrote:

> Reading "ENTITY-ATTRIBUTE AND GRAPH TABLES" in the Table Design page,
> I'm musing about where to put information about information.
>
> If 'A manager_of B' (with confidence 0.5, supported by documents a, b, c)
>
> ...
>
> how would you recommend adding that to the scheme described there?
>

Re: Modeling provenance or other 'facts about facts'

Posted by Billie J Rinaldi <bi...@ugov.gov>.
On Wednesday, February 1, 2012 6:15:33 PM, "Benson Margulies" <bi...@gmail.com> wrote:
> Reading "ENTITY-ATTRIBUTE AND GRAPH TABLES" in the Table Design page,
> I'm musing about where to put information about information.
> 
> If 'A manager_of B' (with confidence 0.5, supported by documents a, b,
> c)
> 
> ...
> 
> how would you recommend adding that to the scheme described there?

I might add the confidence / supporting documents information into the Value in this particular example.  There isn't currently any information in the Value associated with Key E002 manager_of E001.  You could even have a combiner to add new documents as supporting evidence (though the confidence might be more difficult to recalculate).

Billie