Lab: More OWL

2020-04-19T11:03:02Z

Say004:

2020-04-17T10:46:51Z

Say004:

This page will be updated with Python examples related to the lectures and labs. We will add more examples after each lab has ended. The first examples will use Python's RDFlib. We will introduce other relevant libraries later.

==Lecture 1: Python, RDFlib, and PyCharm==

===Printing the triples of the Graph in a readable way===
<syntaxhighlight>
# The turtle format has the purpose of being more readable for humans.
print(g.serialize(format="turtle").decode())
</syntaxhighlight>

===Coding Tasks Lab 1===
<syntaxhighlight>
from rdflib import Graph, Namespace, URIRef, BNode, Literal
from rdflib.namespace import RDF, FOAF, XSD

g = Graph()
ex = Namespace("http://example.org/")

g.add((ex.Cade, ex.married, ex.Mary))
g.add((ex.France, ex.capital, ex.Paris))
g.add((ex.Cade, ex.age, Literal("27", datatype=XSD.integer)))
g.add((ex.Mary, ex.age, Literal("26", datatype=XSD.integer)))
g.add((ex.Mary, ex.interest, ex.Hiking))
g.add((ex.Mary, ex.interest, ex.Chocolate))
g.add((ex.Mary, ex.interest, ex.Biology))
g.add((ex.Mary, RDF.type, ex.Student))
g.add((ex.Paris, RDF.type, ex.City))
g.add((ex.Paris, ex.locatedIn, ex.France))
g.add((ex.Cade, ex.characteristic, ex.Kind))
g.add((ex.Mary, ex.characteristic, ex.Kind))
g.add((ex.Mary, RDF.type, FOAF.Person))
g.add((ex.Cade, RDF.type, FOAF.Person))

</syntaxhighlight>

==Lecture 2: RDF programming==

===Different ways to create an address===

<syntaxhighlight>

from rdflib import Graph, Namespace, URIRef, BNode, Literal
from rdflib.namespace import RDF, FOAF, XSD

g = Graph()
ex = Namespace("http://example.org/")

# How to represent the address of Cade Tracey. From probably the worst solution to the best.

# Solution 1 -
# Make the entire address into one Literal. However, Generally we want to separate each part of an address into their own triples. This is useful for instance if we want to find only the streets where people live.

g.add((ex.Cade_Tracey, ex.livesIn, Literal("1516_Henry_Street, Berkeley, California 94709, USA")))

# Solution 2 -
# Seperate the different pieces information into their own triples

g.add((ex.Cade_tracey, ex.street, Literal("1516_Henry_Street")))
g.add((ex.Cade_tracey, ex.city, Literal("Berkeley")))
g.add((ex.Cade_tracey, ex.state, Literal("California")))
g.add((ex.Cade_tracey, ex.zipcode, Literal("94709")))
g.add((ex.Cade_tracey, ex.country, Literal("USA")))

# Solution 3 - Some parts of the addresses can make more sense to be resources than Literals.
# Larger concepts like a city or state are typically represented as resources rather than Literals, but this is not necesarilly a requirement in the case that you don't intend to say more about them.

g.add((ex.Cade_tracey, ex.street, Literal("1516_Henry_Street")))
g.add((ex.Cade_tracey, ex.city, ex.Berkeley))
g.add((ex.Cade_tracey, ex.state, ex.California))
g.add((ex.Cade_tracey, ex.zipcode, Literal("94709")))
g.add((ex.Cade_tracey, ex.country, ex.USA))

# Solution 4
# Grouping of the information into an Address. We can Represent the address concept with its own URI OR with a Blank Node.
# One advantage of this is that we can easily remove the entire address, instead of removing each individual part of the address.
# Solution 4 or 5 is how I would recommend to make addresses. Here, ex.CadeAddress could also be called something like ex.address1 or so on, if you want to give each address a unique ID.

# Address URI - CadeAdress

g.add((ex.Cade_Tracey, ex.address, ex.CadeAddress))
g.add((ex.CadeAddress, RDF.type, ex.Address))
g.add((ex.CadeAddress, ex.street, Literal("1516 Henry Street")))
g.add((ex.CadeAddress, ex.city, ex.Berkeley))
g.add((ex.CadeAddress, ex.state, ex.California))
g.add((ex.CadeAddress, ex.postalCode, Literal("94709")))
g.add((ex.CadeAddress, ex.country, ex.USA))

# OR

# Blank node for Address.
address = BNode()
g.add((ex.Cade_Tracey, ex.address, address))
g.add((address, RDF.type, ex.Address))
g.add((address, ex.street, Literal("1516 Henry Street", datatype=XSD.string)))
g.add((address, ex.city, ex.Berkeley))
g.add((address, ex.state, ex.California))
g.add((address, ex.postalCode, Literal("94709", datatype=XSD.string)))
g.add((address, ex.country, ex.USA))

# Solution 5 using existing vocabularies for address

# (in this case https://schema.org/PostalAddress from schema.org).
# Also using existing ontology for places like California. (like http://dbpedia.org/resource/California from dbpedia.org)

schema = "https://schema.org/"
dbp = "https://dpbedia.org/resource/"

g.add((ex.Cade_Tracey, schema.address, ex.CadeAddress))
g.add((ex.CadeAddress, RDF.type, schema.PostalAddress))
g.add((ex.CadeAddress, schema.streetAddress, Literal("1516 Henry Street")))
g.add((ex.CadeAddress, schema.addresCity, dbp.Berkeley))
g.add((ex.CadeAddress, schema.addressRegion, dbp.California))
g.add((ex.CadeAddress, schema.postalCode, Literal("94709")))
g.add((ex.CadeAddress, schema.addressCountry, dbp.United_States))

</syntaxhighlight>

===Typed Literals===
<syntaxhighlight>
from rdflib import Graph, Literal, Namespace
from rdflib.namespace import XSD
g = Graph()
ex = Namespace("http://example.org/")

g.add((ex.Cade, ex.age, Literal(27, datatype=XSD.integer)))
g.add((ex.Cade, ex.gpa, Literal(3.3, datatype=XSD.float)))
g.add((ex.Cade, FOAF.name, Literal("Cade Tracey", datatype=XSD.string)))
g.add((ex.Cade, ex.birthday, Literal("2006-01-01", datatype=XSD.date)))
</syntaxhighlight>

===Writing and reading graphs/files===

<syntaxhighlight>
# Writing the graph to a file on your system. Possible formats = turtle, n3, xml, nt.
g.serialize(destination="triples.txt", format="turtle")

# Parsing a local file
parsed_graph = g.parse(location="triples.txt", format="turtle")

# Parsing a remote endpoint like Dbpedia
dbpedia_graph = g.parse("http://dbpedia.org/resource/Pluto")
</syntaxhighlight>

===Collection Example===

<syntaxhighlight>
from rdflib import Graph, Namespace
from rdflib.collection import Collection

# Sometimes we want to add many objects or subjects for the same predicate at once.
# In these cases we can use Collection() to save some time.
# In this case I want to add all countries that Emma has visited at once.

b = BNode()
g.add((ex.Emma, ex.visit, b))
Collection(g, b,
[ex.Portugal, ex.Italy, ex.France, ex.Germany, ex.Denmark, ex.Sweden])

# OR

g.add((ex.Emma, ex.visit, ex.EmmaVisits))
Collection(g, ex.EmmaVisits,
[ex.Portugal, ex.Italy, ex.France, ex.Germany, ex.Denmark, ex.Sweden])

</syntaxhighlight>

==Lecture 3: SPARQL==

===SPARQL queries from the lecture===
<syntaxhighlight>
SELECT DISTINCT ?p WHERE {
?s ?p ?o .
}
</syntaxhighlight>

<syntaxhighlight>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

SELECT DISTINCT ?t WHERE {
?s rdf:type ?t .
}
</syntaxhighlight>

<syntaxhighlight>
PREFIX owl: <http://www.w3.org/2002/07/owl#>
CONSTRUCT {
?s owl:sameAs ?o2 .
} WHERE {
?s owl:sameAs ?o .
FILTER(REGEX(STR(?o), "^http://www\\.", "s"))
BIND(URI(REPLACE(STR(?o), "^http://www\\.", "http://", "s")) AS ?o2)
}
</syntaxhighlight>

===Select all contents of lists (rdfllib.Collection)===
<syntaxhighlight>

# rdflib.Collection has a different interntal structure so it requires a slightly more advance query. Here I am selecting all places that Emma has visited.

PREFIX ex: <http://example.org/>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

SELECT ?visit
WHERE {
ex:Emma ex:visit/rdf:rest*/rdf:first ?visit
}
</syntaxhighlight>

==Lecture 4- SPARQL PROGRAMMING==

===SELECTING data from Blazegraph via Python===
<syntaxhighlight>

from SPARQLWrapper import SPARQLWrapper, JSON

# This creates a server connection to the same URL that contains the graphic interface for Blazegraph.
# You also need to add "sparql" to end of the URL like below.

sparql = SPARQLWrapper("http://localhost:9999/blazegraph/sparql")

# SELECT all triples in the database.

sparql.setQuery("""
SELECT DISTINCT ?p WHERE {
?s ?p ?o.
}
""")
sparql.setReturnFormat(JSON)
results = sparql.query().convert()

for result in results["results"]["bindings"]:
print(result["p"]["value"])

# SELECT all interests of Cade

sparql.setQuery("""
PREFIX ex: <http://example.org/>
SELECT DISTINCT ?interest WHERE {
ex:Cade ex:interest ?interest.
}
""")
sparql.setReturnFormat(JSON)
results = sparql.query().convert()

for result in results["results"]["bindings"]:
print(result["interest"]["value"])
</syntaxhighlight>

===Updating data from Blazegraph via Python===
<syntaxhighlight>
from SPARQLWrapper import SPARQLWrapper, POST, DIGEST

namespace = "kb"
sparql = SPARQLWrapper("http://localhost:9999/blazegraph/namespace/"+ namespace + "/sparql")

sparql.setMethod(POST)
sparql.setQuery("""
PREFIX ex: <http://example.org/>
INSERT DATA{
ex:Cade ex:interest ex:Mathematics.
}
""")

results = sparql.query()
print(results.response.read())

</syntaxhighlight>

== Lecture 5: RDFS==

===RDFS inference with RDFLib===
You can use the OWL-RL package to add inference capabilities to RDFLib. Download it [https://github.com/RDFLib/OWL-RL GitHub] and copy the ''owlrl'' subfolder into your project folder next to your Python files.

[https://owl-rl.readthedocs.io/en/latest/owlrl.html OWL-RL documentation.]

Example program to get started:
<syntaxhighlight>
import rdflib.plugins.sparql.update
import owlrl.RDFSClosure

g = rdflib.Graph()

ex = rdflib.Namespace('http://example.org#')
g.bind('', ex)

g.update("""
PREFIX ex: <http://example.org#>
PREFIX owl: <http://www.w3.org/2002/07/owl#>
INSERT DATA {
ex:Socrates rdf:type ex:Man .
ex:Man rdfs:subClassOf ex:Mortal .
}""")

# The next three lines add inferred triples to g.
rdfs = owlrl.RDFSClosure.RDFS_Semantics(g, False, False, False)
rdfs.closure()
rdfs.flush_stored_triples()

b = g.query("""
PREFIX ex: <http://example.org#>
ASK {
ex:Socrates rdf:type ex:Mortal .
}
""")
print('Result: ' + bool(b))
</syntaxhighlight>

===Languaged tagged RDFS labels===
<syntaxhighlight>
from rdflib import Graph, Namespace, Literal
from rdflib.namespace import RDFS

g = Graph()
ex = Namespace("http://example.org/")

g.add((ex.France, RDFS.label, Literal("Frankrike", lang="no")))
g.add((ex.France, RDFS.label, Literal("France", lang="en")))
g.add((ex.France, RDFS.label, Literal("Francia", lang="es")))

</syntaxhighlight>

== Lecture 6: RDFS Plus / OWL ==
===RDFS Plus / OWL inference with RDFLib===

You can use the OWL-RL package again as for Lecture 5.

Instead of:
<syntaxhighlight>
# The next three lines add inferred triples to g.
rdfs = owlrl.RDFSClosure.RDFS_Semantics(g, False, False, False)
rdfs.closure()
rdfs.flush_stored_triples()
</syntaxhighlight>
you can write this to get both RDFS and basic RDFS Plus / OWL inference:
<syntaxhighlight>
# The next three lines add inferred triples to g.
owl = owlrl.CombinedClosure.RDFS_OWLRL_Semantics(g, False, False, False)
owl.closure()
owl.flush_stored_triples()
</syntaxhighlight>

Example updates and queries:
<syntaxhighlight>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
PREFIX owl: <http://www.w3.org/2002/07/owl#>
PREFIX ex: <http://example.org#>

INSERT DATA {
ex:Socrates ex:hasWife ex:Xanthippe .
ex:hasHusband owl:inverseOf ex:hasWife .
}
</syntaxhighlight>

<syntaxhighlight>
ASK {
ex:Xanthippe ex:hasHusband ex:Socrates .
}
</syntaxhighlight>

<syntaxhighlight>
ASK {
ex:Socrates ^ex:hasHusband ex:Xanthippe .
}
</syntaxhighlight>

<syntaxhighlight>
INSERT DATA {
ex:hasWife rdfs:subPropertyOf ex:hasSpouse .
ex:hasSpouse rdf:type owl:SymmetricProperty .
}
</syntaxhighlight>

<syntaxhighlight>
ASK {
ex:Socrates ex:hasSpouse ex:Xanthippe .
}
</syntaxhighlight>

<syntaxhighlight>
ASK {
ex:Socrates ^ex:hasSpouse ex:Xanthippe .
}
</syntaxhighlight>

== Lab 9 ==

===Download from BlazeGraph===

<syntaxhighlight>
"""
Dumps a database to a local RDF file.
You need to install the SPARQLWrapper package first...
"""

import datetime
from SPARQLWrapper import SPARQLWrapper, RDFXML

# your namespace, the default is 'kb'
ns = 'kb'

# the SPARQL endpoint
endpoint = 'http://info216.i2s.uib.no/bigdata/namespace/' + ns + '/sparql'

# - the endpoint just moved, the old one was:
# endpoint = 'http://i2s.uib.no:8888/bigdata/namespace/' + ns + '/sparql'

# create wrapper
wrapper = SPARQLWrapper(endpoint)

# prepare the SPARQL update
wrapper.setQuery('CONSTRUCT { ?s ?p ?o } WHERE { ?s ?p ?o }')
wrapper.setReturnFormat(RDFXML)

# execute the SPARQL update and convert the result to an rdflib.Graph
graph = wrapper.query().convert()

# the destination file, with code to make it timestamped
destfile = 'rdf_dumps/slr-kg4news-' + datetime.datetime.now().strftime('%Y%m%d-%H%M') + '.rdf'

# serialize the result to file
graph.serialize(destination=destfile, format='ttl')

# report and quit
print('Wrote %u triples to file %s .' %
(len(res), destfile))
</syntaxhighlight>

==Semantic Lifting - CSV==

<syntaxhighlight>
from rdflib import Graph, Literal, Namespace, URIRef
from rdflib.namespace import RDF, FOAF, RDFS, OWL
import pandas as pd

g = Graph()
ex = Namespace("http://example.org/")
g.bind("ex", ex)

# Load the CSV data as a pandas Dataframe.
csv_data = pd.read_csv("task1.csv")

# Here I deal with spaces (" ") in the data. I replace them with "_" so that URI's become valid.
csv_data = csv_data.replace(to_replace=" ", value="_", regex=True)

# Here I mark all missing/empty data as "unknown". This makes it easy to delete triples containing this later.
csv_data = csv_data.fillna("unknown")

# Loop through the CSV data, and then make RDF triples.
for index, row in csv_data.iterrows():
# The names of the people act as subjects.
subject = row['Name']
# Create triples: e.g. "Cade_Tracey - age - 27"
g.add((URIRef(ex + subject), URIRef(ex + "age"), Literal(row["Age"])))
g.add((URIRef(ex + subject), URIRef(ex + "married"), URIRef(ex + row["Spouse"])))
g.add((URIRef(ex + subject), URIRef(ex + "country"), URIRef(ex + row["Country"])))

# If We want can add additional RDF/RDFS/OWL information e.g
g.add((URIRef(ex + subject), RDF.type, FOAF.Person))

# I remove triples that I marked as unknown earlier.
g.remove((None, None, URIRef("http://example.org/unknown")))

# Clean printing of the graph.
print(g.serialize(format="turtle").decode())
</syntaxhighlight>

===CSV file for above example===

<syntaxhighlight>
"Name","Age","Spouse","Country"
"Cade Tracey","26","Mary Jackson","US"
"Bob Johnson","21","","Canada"
"Mary Jackson","25","","France"
"Phil Philips","32","Catherine Smith","Japan"
</syntaxhighlight>

==Semantic Lifting - XML==
<syntaxhighlight>
from rdflib import Graph, Literal, Namespace, URIRef
from rdflib.namespace import RDF, XSD, RDFS
import xml.etree.ElementTree as ET

g = Graph()
ex = Namespace("http://example.org/TV/")
prov = Namespace("http://www.w3.org/ns/prov#")
g.bind("ex", ex)
g.bind("prov", prov)

tree = ET.parse("tv_shows.xml")
root = tree.getroot()

for tv_show in root.findall('tv_show'):
show_id = tv_show.attrib["id"]
title = tv_show.find("title").text

g.add((URIRef(ex + show_id), ex.title, Literal(title, datatype=XSD.string)))
g.add((URIRef(ex + show_id), RDF.type, ex.TV_Show))

for actor in tv_show.findall("actor"):
first_name = actor.find("firstname").text
last_name = actor.find("lastname").text
full_name = first_name + "_" + last_name

g.add((URIRef(ex + full_name), ex.starsIn, URIRef(title)))
g.add((URIRef(ex + full_name), RDF.type, ex.Actor))

print(g.serialize(format="turtle").decode())
</syntaxhighlight>

===XML Data for above example===
<syntaxhighlight>
<data>
<tv_show id="1050">
<title>The_Sopranos</title>
<actor>
<firstname>James</firstname>
<lastname>Gandolfini</lastname>
</actor>
</tv_show>
<tv_show id="1066">
<title>Seinfeld</title>
<actor>
<firstname>Jerry</firstname>
<lastname>Seinfeld</lastname>
</actor>
<actor>
<firstname>Julia</firstname>
<lastname>Louis-dreyfus</lastname>
</actor>
<actor>
<firstname>Jason</firstname>
<lastname>Alexander</lastname>
</actor>
</tv_show>
</data>
</syntaxhighlight>

==WEB API Calls (In this case JSON)==
<syntaxhighlight>
import requests
import json
import pprint

# Retrieve JSON data from API service URL. Then load it with the json library as a json object.
url = "http://api.geonames.org/postalCodeLookupJSON?postalcode=46020&#country=ES&username=demo"
data = requests.get(url).content.decode("utf-8")
data = json.loads(data)
pprint.pprint(data)
</syntaxhighlight>

==JSON-LD==

<syntaxhighlight>
import rdflib

g = rdflib.Graph()

example = """
{
"@context": {
"name": "http://xmlns.com/foaf/0.1/name",
"homepage": {
"@id": "http://xmlns.com/foaf/0.1/homepage",
"@type": "@id"
}
},
"@id": "http://me.markus-lanthaler.com/",
"name": "Markus Lanthaler",
"homepage": "http://www.markus-lanthaler.com/"
}
"""

# json-ld parsing automatically deals with @contexts
g.parse(data=example, format='json-ld')

# serialisation does expansion by default
for line in g.serialize(format='json-ld').decode().splitlines():
print(line)

# by supplying a context object, serialisation can do compaction
context = {
"foaf": "http://xmlns.com/foaf/0.1/"
}
for line in g.serialize(format='json-ld', context=context).decode().splitlines():
print(line)
</syntaxhighlight>

<div class="credits" style="text-align: right; direction: ltr; margin-left: 1em;">''INFO216, UiB, 2017-2020. All code examples are [https://creativecommons.org/choose/zero/ CC0].'' </div>

Lab Solutions

2020-04-17T10:10:30Z

Say004:

2020-04-17T07:41:39Z

Say004:

=Lab 12: Accessing and lifting Web APIs (RESTful web services)=

==Topics==
Programming regular (non-semantic) as well as semantic Web APIs (RESTful web services) with JSON and JSON-LD.
We will use Web APIs to retrieve regular JSON data, and then append it with a semantic context (@context).
Finally we will parse it with RDFlib.

==Imports==
* import requests
* import json
* import pprint
* from rdflib import Graph

==Tasks==
===Regular JSON web APIs===
Write a small program that accesses a regular (non-semantic) web API and download the result. The "json" library in python can be used to load a json string as a json object (json.loads(data)).
Use the the prettyprint import to print a readable version of the json object.

The GeoNames web API (http://www.geonames.org/export/ws-overview.html) offers many services. For example, you can use this URL to access more information about Ines' neighbourhood in Valencia: http://api.geonames.org/postalCodeLookupJSON?postalcode=46020&country=ES&username=demo (You might need to register a username instead of using "demo"). '
You can register here if you want to: https://www.geonames.org/login

You do not have to use the GeoNames web API. There are lots and lots of other web APIs out there. But we preferably want something simple that does not require extensive registration (HTTPS can also make things more complex when the certificates are outdated). Here are some examples to get you started if you want to try out other APIs: http://opendata.app.uib.no/ , http://data.ssb.no/api , http://ws.audioscrobbler.com/2.0/ , http://www.last.fm/api /intro , http://wiki.musicbrainz.org/Development/JSON_Web_Service .

While you are testing and debugging things, it is good to make measures so that you do not need to call the GeoNames or other API over and over. A solution can be writing the returned data to a file, or copying it into a variable.

Here is an example of a results string you can use, if you have trouble connecting to GeoNames (note that you have to escape all the quotation marks inside the Java string):
{\"postalcodes\":[{\"adminCode2\":\"V\",\"adminCode1\":\"VC\",\"adminName2\":\"Valencia\",\"lng\":-0.377386808395386,\"countryCode\":\"ES\",\"postalcode\":\"46020\",\"adminName1\":\"Comunidad Valenciana\",\"placeName\":\"Valencia\",\"lat\":39.4697524227712}]}"

===Lifting JSON to JSON-LD===

In python we can represent JSON objects as dictionaries ({}) and JSON Arrays as lists ([]).

So far we have only used plain JSON. Now we want to move to JSON-LD, the semantic version of JSON. Make a new JSON object (dictionary/{} in python) called context. Put a single entry into this map, with "@context" as the key and another object ({}) as the value. It is this second map that contains the actual mappings.

Put at least one pair of strings into it. For example, if you used the postcode API, the pair "lat" and "http://www.w3.org/2003/01/geo/wgs84_pos#lat". You can also put the pair "lng" and "http://www.w3.org/2003/01/geo/wgs84_pos#long".

Add this pair too to the context object: "postalcodes" and "http://dbpedia.org/ontology/postalCode".

Add more string pairs, using existing or inventing new terms as you go along, to the context object.

Go back to the RDF/RDFS programs your wrote in labs 2 and 3 if you can. Extend the program so that it adds further information about the post codes of every person in your graph.

We will now make a RDFlib Graph from the JSON-LD object.

First you need to pip install the json-ld portion of rdflib if you have not already:
<syntaxhighlight>
pip install rdflib-jsonld
</syntaxhighlight>

Now, create a new Graph. Then convert the JSON-LD object to a string (use json.dumps() and write it to a file). Then parse the file with Rdflib (g.parse()).

Congratulations - you have now gone through the steps of accessing a web API over the net, lifting the results using JSON-LD, manipulating the in JSON-LD and reading them into a RDF Graph. Of course, it is easy to convert the RDFlib graph back into JSON-LD using g.serialize("json-ld")

===If You have more time===

===Useful Reading===
[https://stackabuse.com/reading-and-writing-json-to-a-file-in-python/ - Reading and writing with JSON - stackabuse.com]