[rdf4j-dev] shac, statistics on classes etc

["Bart Hanssens (BOSA)" <bart.hanssens@xxxxxxxxxxxx>]

Hi,

 

Just a quick note and some thoughts.

 

I’m developing a stand-alone SHACL validator, nothing fancy, which is to be integrated in my data.gov.be toolchain

https://github.com/Fedict/shaclvalidator

 

Of course the SHACL part works like charm, thanks Håvard 😉

Only a few minor issues that will either be solved in 4.3 (severity level),

or are arguably issues with the SHACL files on semic.eu (name on nodeshape, and nodeshapes with empty shacl:property)

 

I was wondering if it would be hard (or interesting for other people) to collect statistics on

1. number of times a shape did _not_ have validation issues , or how many times a shape matched in total
2. number of different classes/properties/object values in a dataset

 

Use case for (a) is mainly a metric for data quality (shape violations divided by total),

while (b) is useful for harmonizing data (eg reducing differences) but probably useful for optimizing queries / data storage as well.

 

For the time being I’m (ab)using data cubes for publishing the statistics in TTL, but perhaps there is a better vocabulary.

And I’m guessing some data stores already collect some of this data.

 

Happy to look into it myself, though hints on how to get started would be appreciated 😊

 

 

Best regards,

 

Bart