how best to create/work-with a linkset or tagset of an arbitrary number of relationship predicates between two lists of objects

[Simon Opper]

  Hi folks

I want to encode relationships between two list of things,   a list of agencies and a list of functions.  The relationships are defined in a third list.

As one outcome, I also want to test the ability of Maui to attribute relationships. I think it will be poor but I want to confirm.

A tagset doesn't work out of the box as it just has one list of things that can be tagged with something. I could either tag agencies with a relationship... but to what. Or the agency to a function, but only with a single tag predicate e.g skos:related. So I still then  have to attribute the relationships somehow. 

A crosswalk doesn't work out of the box as while you can have both the set of agencies and the functions, it only allows a single match predicate e.g. crosswlk:closeMatch  and I want to be able to select from a range of possible match predicates.

So I can obviously work through this through in an ontology or manual code editing manner to derive such relations but I'm hoping that I can leverage the existing asset interfaces and maui:

can the tagger can be adapted to enable a dynamic set of tag predicates and not just  a single skos property for example. possibly a shacl shape  or rule.  Before I try this I just wanted to see if the tagger interface and maui would still work ?

and/or could the crosswalks asset be customised in a similar manner to allow a dynamic list of match predicates ?

Many thanks in advance

Simon

[Irene Polikoff]

You can use multiple relationships in a tagset, but a list of relationships is static I.e., the choice of relationships is not dependent on the items you are connecting. As I understand it, you are wanting different relationships depending on the type of items you are connecting.

Further, auto classification will always use a single relationship. What we typically do is run a set of rules, post classification, to transform this single relationship into different relationships reflecting the type of connected items. For example, if there is a tagging link between an agency and a function, a rule could generate ex:activity link between them. 

Having said this, the auto classification algorithm is about tagging documents (content)  with the most relevant concepts, based on the document text. I am not sure where is the content in the agency and function example. Rule processing I have described is typically used where tagging concepts are of different type and a relationship between a document and a tag would differ, depending on the type of a tag e.g., if it is a Geo Concept, relationship may be ‘location’, if it is a business, relationship may be ‘organization’, etc.

Once generated, you will be able to easily see these links in the default view of the Tagset editor if these relationships are amongst the possible  tagging relationships. Otherwise, you would need to use a different view to see them.

Similarly, for manual tagging, you will be able to use these relationships if they are on the list of possible tagging relationships.

For crosswalks, the links created through the built-in auto-mapping algorithm will always use a single selected predicate. There is a way to plug in custom mapping algorithms. The default UI (crosswalk table), will only show side by side items connected by the currently selected mapping predicate. Again, you would need to use an alternative layout to see items connected by any predicate.

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/topbraid-users/5b2daa7b-9542-40f7-b65e-56c50bf4f70bn%40googlegroups.com.