Normalizing schema.org namespace

Consistent representation of namespaces simplifies programmatic processing of markup. For example, even though conceptually it is clear the terms http://schema.org/Dataset and https://schema.org/Dataset (note the protocol difference) are referring to https://schema.org/Dataset, these are programmatically treated as different entities. The schema.org guidelines [1] are somewhat ambivalent on the topic, with some emphasis on “https” adoption.

The trailing slash (/) is also important. Without it, common RDF processing libraries such as rdflib [2] will construct a term like “https://schema.orgDataset”. For example given the three SO:Dataset entries:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
[
  {
    "@context": {
      "@vocab": "https://schema.org"
    },
    "@id":"demo_0",
    "@type":"Dataset",
    "name": "https, no trailing slash"
  },
  {
    "@context": {
      "@vocab": "http://schema.org"
    },
    "@id":"demo_1",
    "@type":"Dataset",
    "name": "http, no trailing slash"
  },
  {
    "@context": {
      "@vocab": "http://schema.org/"
    },
    "@id":"demo_2",
    "@type":"Dataset",
    "name": "http only"
  }
]

the resulting RDF:type values as recognized by rdflib are:

from rdflib import ConjunctiveGraph
g = ConjunctiveGraph()
g.parse(
    "examples/data/ds_bad_namespace.json",
    format="json-ld",
    publicID="https://my.server.net/data/"
)
sparql_types = """
  PREFIX rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
  SELECT ?x ?y
  {
     ?x rdf:type ?y .
  } ORDER BY ASC(?x) """

qres = g.query(sparql_types)
for v in qres:
    print(f'{v[0].n3()} {v[1].n3()}')

<https://my.server.net/data/demo_0> <https://schema.orgDataset/>
<https://my.server.net/data/demo_1> <http://schema.orgDataset/>
<https://my.server.net/data/demo_2> <http://schema.org/Dataset>

None of these will match for a query that relies on the schema.org namespace of https://schema.org/:

sparql = """
  PREFIX rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
  PREFIX SO: <https://schema.org/>
  SELECT ?x
  {
     ?x rdf:type SO:Dataset .
  } """

qres = g.query(sparql)
print(f"Number of matches = {len(qres)}")

Number of matches = 0

There are however, many real world examples of each use of the schema.org namespace variants.

In order to simplify tooling for extracting information from schema.org resources, sotools will normalize all json-ld graphs to use the namespace https://schema.org/. The normalization is performed when loading with the sotools.common.loadSOGraph() method. For example, loading the same data with normalization and applying the same queries:

import sotools

g = sotools.loadSOGraph(
    filename="examples/data/ds_bad_namespace.json",
    publicID="https://my.server.net/data/"
)

qres = g.query(sparql_types)
for v in qres:
    print(f'{v[0].n3()} {v[1].n3()}')

qres = g.query(sparql)
print(f"Number of matches = {len(qres)}")

<https://my.server.net/data/demo_0> <https://schema.org/Dataset>
<https://my.server.net/data/demo_1> <https://schema.org/Dataset>
<https://my.server.net/data/demo_2> <https://schema.org/Dataset>
Number of matches = 3

Running code on this page

All examples on this page can be run live in Binder. To do so:

  1. Click on the “Activate Binder” button
  2. Wait for Binder to be active. This can take a while, you can watch progress in your browser’s javascript console. When a line like Kernel: connected (89dfd3c8... appears, Binder should be ready to go.
  3. Run the following before any other script on the page. This sets the right path context for loading examples etc.
import os
try:
    os.chdir("docsource/source")
except:
    pass
print("Page is ready. You can now run other code blocks on this page.")