Re: Why is PhysicalThing.additionalType recommended rather than mandatory

Thanks for making that change, Leyla.  I wasn't sure about the exact status
of the PhysicalEntity/Record specifications.  It would be very good to go
through them as a group in October.  Personally, I like this new structure
from what I've seen so far.

In the InterMine case it might be that a user can't provide a URI because
their new class is something that isn't in an ontology (yet).  Initially, I
was thinking that maybe we could allow a bare string, but I also like
Michel's suggestion that it could be a link to a general term.  However, I
wonder if this could affect findability, e.g. an InterMine user labels an
entity as an 'organic polymer' [1] but it gets swamped in a search engine
because there are also millions of entities in organic polymer subclasses.

Or maybe even this minimum level of structure, when combined with a bit
more information like a name term, is sufficient to whittle down the hits
to make cross data source search more useful than at present.

[1] http://semanticscience.org/resource/SIO_010346.rdf

On 19 September 2017 at 13:36, Leyla Garcia <ljgarcia@ebi.ac.uk> wrote:

> On 19/09/2017 12:35, Justin Clark-Casey wrote:
>
>> I see that PhysicalThing.additionalType is shown as recommended [1]
>> whereas BiologicalEntity.biologicalType [2] was mandatory.  What is the
>> reason for this change?  I thought that this was one for the most critical
>> properties (since most entity types will not have their own Bioschemas
>> subclass).
>>
>> [1] http://bioschemas.org/bsc_specs/PhysicalEntity/specification/
>> [2] https://docs.google.com/document/d/1XASuESIHU3bi1aXMxQS5-rCO
>> QX0ugjMNkh68VF4co4Q
>>
>> -- Justin
>>
>
> I might be wrong but I think the specifications are still work in
> progress. I am taking your comment as a suggestion for PhysicalEntity.
> Still, we should take a second look to M/R/O for PhysicalEntity and Record
> and get to some agreements as a group. I think the point you raised makes
> sense, additionalType should be mandatory (and it has been modified as
> such). However, I am wondering if anyone has a case where the biological
> type cannot be provided. Also, is ONE enough? Should be allowed MANY for
> that field? And, if we allow many, will sameAs be assumed by applications?
>
> Regards,
>
>