The reporting has been nice for understanding the extent and frequency of properties to begin understanding what a core description might be for different entity types when reusing the data, not just a list of the properties *ever* used in the dataset.
There was also a point when we were considering creating FAST-like local entities and wanted to be sure we were using similar semantics. Now, as we consider creating BF and related RWOs/Authorities with practices aligned with LC, the reporting could inform profile development. Linking to related profiles would be nice too. :)
Hope this helps,
On 11/13/18, 4:25 PM, "LC Linked Data Service Discussion List on behalf of Ford, Kevin" <[log in to unmask] on behalf of [log in to unmask]> wrote:
As to where the documentation for these types of decisions live, I'm not 100% sure I can answer the question, but only because, until now (and technically we've made no decision yet), there should be no difference between the bulk downloads and what you see from the web-based service to warrant documentation. That said, if you look at the top of the "Downloads" page  you'll find a short paragraph and a link to additional information. I suspect we will expand that short paragraph to make sure notice about this difference is actually *on* the downloads page, and then expand on it elsewhere, which is where users would find the SPARQL samples. Perhaps, if we really get elaborate, we make a gist or a small repo in which to store the SPARQL queries, but I'm speculating. Other than committing to providing the needed SPARQL queries, we've not identified a how or a where.
As for the FAST statistics, how do these help you? How do you use the resulting property and class reports? We do not currently generate this information but that is not to say we couldn't, especially if there is use for them beyond our walls. Do the counts matter or is it about seeing a list of used properties and classes in the system?
From: LC Linked Data Service Discussion List <[log in to unmask]> On Behalf Of Steven Folsom
Sent: Tuesday, November 13, 2018 3:19 PM
To: [log in to unmask]
Subject: Re: [ID.LOC.GOV] Derived relationships in bulk downloads, a question
Seems pretty reasonable from the downloads perspective for the reasons you gave, as long as the documentation is clear and all the semantics are queryable from one direction or the other. Generally, it seems more important to have the fuller representation of specific entities in the web based descriptions so that when dereferenced we have a complete view.
Where does the documentation for these types of decisions currently live? I couldn't find anything on ID's site; this doesn't mean it isn't there. :) Related, have you considered providing statistics like FAST does? http://experimental.worldcat.org/fast/stats/FASTLinkedDataProfile.html (I realize you might end up having different statistics for the download and the web based descriptions, but I find this reporting really useful instead of having to query myself to create property and class reports.)
Thanks for your work on this. It's exciting to think we might see more frequent data dumps.