lcnetdev / lc-bulk-export-pilot Goto Github PK

Don't copy the context in every line. Put it on the web and reference it as a URL.

This will save significant megabytes. all modern clients cache network contexts.

don't emit ^^xsd:string, it's the default literal datatype.

http://id.loc.gov/authorities/subjects/sh00000011.json
You seem to include all related nodes in the payload (thanks!)

Can you also describe http://id.loc.gov/vocabulary/organizations/dlc ?

Both VOID (eg http://id.loc.gov/static/data/downloads/lcsh.skos.ndjson.json) and data files (eg http://id.loc.gov/authorities/subjects/sh00000011.json) are JSONLD not mere JSON, so:

• use jsonld extension. Eg otherwise I need to tell Jena RIOT what the file is
• use `application/ld+json'

You can of course also keep the old ones for backward compatibility.

Update http://id.loc.gov/techcenter/serializations.html and http://id.loc.gov/techcenter/downloads.html

http://id.loc.gov/static/data/downloads/lcsh.skos.ndjson.json is rather poor:
title, description, date and hash.

See eg http://vocab.getty.edu/doc/#Descriptive_Information for some very detailed VOID, and a way to mix VOID and DCAT. DataID is also worth exploring.

• It doesn't say it's related to the data file (http://id.loc.gov/static/data/downloads/lcsh.skos.ndjson) in any way
• It needs to link to all (content-negotiated) representations of the data file
• use dct:modified not dct:date
• state as a minimum void:entities, void:triples, dc:format, dcat:byteSize
• pick some more http://vocab.getty.edu/doc/#Descriptive_Properties to add, eg license, void:exampleResource etc

Why are you getting rid of Turtle? It's much more readable and compact than NT.
Even if you emit line-oriented Turtle (one triple per line, no subject and subject-prefix clusters), this will leverage prefixes to shorten the URLs significantly.

Turtle is as splittable as NT: you can concat any number of turtles together, repeating prefixes freely.

http://id.loc.gov/authorities/subjects/sh00000011.json

• provide some context to shorten the URLs. Otherwise the prop names in jsonld and turtle are not readable. I've collected the following (in turtle)

@prefix bf:       <http://id.loc.gov/ontologies/bibframe/> .
@prefix bflc:     <http://id.loc.gov/ontologies/bflc/> .
@prefix cs:       <http://purl.org/vocab/changeset/schema#> .
@prefix dct:      <http://purl.org/dc/terms/> .
@prefix lc-lang:  <http://id.loc.gov/vocabulary/iso639-2/> .
@prefix lc-org:   <http://id.loc.gov/vocabulary/organizations/> .
@prefix lc-rec:   <http://id.loc.gov/ontologies/RecordInfo#> .
@prefix lc-works: <http://id.loc.gov/resources/works/> .
@prefix lcsh:     <http://id.loc.gov/authorities/subjects/> .
@prefix mads:     <http://www.loc.gov/mads/rdf/v1#> .
@prefix owl:      <http://www.w3.org/2002/07/owl#> .
@prefix rdfs:     <http://www.w3.org/2000/01/rdf-schema#> .
@prefix skos:     <http://www.w3.org/2004/02/skos/core#> .
@prefix xsd:      <http://www.w3.org/2001/XMLSchema#> .