PyCon 2009 IISc, Bangalore, India
Semantic Web and Python
Concepts to Application development
Vinay Modi
Voice Pitara Technologies Private Limited
Outline
Web Need better web for the future Knowledge Representation (KR) to Web Challenges Data integration challenges KR to Web - solutions for challenges Metadata and Semantic Web protocol stack RDF, RDFS and SPARQL basic concepts Using RDFLib adding triples RDFLib serialization RDFLib RDFS ontology Blank node SPARQL querying Graph merging Some possible things one can do with RDFLib
Text in Natural Languages Multimedia Images
Web
Deduce the facts; create mental relationships
Need better Web for the future
I Know What You Mean
KR to Web Challenges
Traditional KR techniques and Network effect Algorithmic complexity and Performance for information space like W3
Scaling KR
KR to Web Challenges
Continue 1
Representational Inconsistencies
Machine down Partial Information
Data integration - Challenges
Web pages, Corporate databases, Institutions Different content and structure Manage for
Company mergers Inter department data sharing (like eGovernment) Research activities/output across labs/nations
Accessible from the web but not public.
Data Integration Challenges
Continue 1
Example: Social sites
add your contacts every time.
Requires standard so that applications can work autonomously and collaboratively.
What is needed
Some data should be available for machines for further processing Data should be possibly combined, merged on Web scale Some time data may describe other data i.e. metadata. Some times data needs to be exchanged. E.g. between Travel preferences and Ticket booking.
Metadata
Data about data Two ways of associating with a resource
Physical embedding Separate resource
Resource identifier Globally unique identifier Advantages of explicit metadata Dublin core, FOAF
KR to Web Solution for Challenges
Extra-logical infrastructure. Network effect Solve syntactic interoperability. Standards
Continue 2
Scalable Representation languages
Semantic Web
Use Web Infrastructure
Semantic Web
Web extension
Exchange Integrate Process Machine automated
Information
RDF basic concepts
W3C decided to build infrastructure for allowing people to make their own vocabularies for talking about different objects. RDF data model:
Resource
Property Literal value
Resource
Property
Resource
RDF basic concepts
Continue 1
RDF graphs and triples:
Subject
http://in.pycon.org/s media/slides/semant icweb_Python.pdf
Predicate title
Object
Semantic Web and Python
RDF Syntax (N3 format):
@prefix dc: <http://http://purl.org/dc/elements/1.1/> .
<http://in.pycon.org/smedia/slides/semanticweb_Pyt hon.pdf> dc:title Semantic Web and Python
RDF basic concepts
Continue 2
Subject (URI)
Predicate (Namespace URI) Object (URI or Literal) Blank Node (Anonymous node; unique to boundary of the domain)
AddisonWesley http://.../isbn/ 67239786
a:publisher
Boston
RDF basic concepts
Continue 3
Ground assertions only. No semantic constraints
Can make anomalous statements
RDFS basic concepts
Extending RDF to make constraints Allows to represent extra-knowledge:
define the terms we can use define the restrictions What other relationships exist
Ontologies
RDFS basic concepts
Continue 1
Classes Instances Sub Classes Properties Sub properties Domain Range
SPARQL basic concepts
Data @prefix foaf: <http://xmlns.com/foaf/0.1/> . _:a foaf:name Vinay" . _:b foaf:name Hari" . Query PREFIX foaf: <http://xmlns.com/foaf/0.1/> SELECT ?name WHERE { ?x foaf:name ?name . } Results (as Python List) [Vinay", Hari"]
SPARQL basic concepts
Query matches the graph:
find a set of variable -> value bindings, such that result of replacing variables by values is a triple in the graph.
SELECT (find values for the given variable and constraint) CONSTRUCT (build a new graph by inserting new values in a triple pattern) ASK (Asks whether a query has a solution in a graph)
RDFLib
Contains Parsers and Serializes for various RDF syntax formats In memory and persistent graph backend RDFLib graphs emulate Python container types best thought of a 3-item triples. [(subject, object, predicate), (subject, object, predicate), ] Ordinary set operations; e.g. add a triple, methods to search triples and return in arbitrary order
RDFLib Adding triple to a graph
from rdflib.Graph import Graph from rdflib import URIRef, Namespace inPyconSlides = Namespace(''http://in.pycon.org/smedia/slides/'') dc = Namespace("http://purl.org/dc/elements/1.1/") g = Graph() g.add((inPyconSlides['Semanticweb_Python.pdf'], dc:title, Literal('Semantic Web and Python concepts to application development')
RDFLib adding triple by reading file/string
str = '''@prefix dc: <''' + dc + '''> . @prefix inPyconSlides : <''' + inPyconSlides + '''> . inPyconSlides :'Semanticweb_Python' dc:title 'Semantic Web and Python concepts to application development' . ''' from rdflib import StringInputSource rdfstr = StringInputSource(str) g.parse(rdfstr, format='n3')
RDFLib adding triple from a remote document
inPyconSlides _rdf = 'http://in.pycon.org/rdf_files/slides.rdf' g.parse(inPyconSlides_rdf, format='n3')
Creating RDFS ontology
Ontology reuse
<http://in.pycon.org> rdf:type <http://swrc.ontoware.org/ ontology#conference> . <http://in.pycon.org/hasSlidesAt> rdf:type rdfs:Property .
<http://in.pycon.org> rdfs:label 'Python Conference, India'
RDFLib SPARQL query
Querying graph instance
# using previous rdf triples q = '''PREFIX dc: <http://purl.org/rss/1.0/> PREFIX inPyconSlides : <http://in.pycon.org/smedia/slides/> SELECT ?x ?y Unbound symbols WHERE { ?x dc:title ?y . } Graph ''' pattern result = g.query(q).serialize(format='n3')
RDFLib creating BNode
from rdflib import BNode profilebnode = BNode()
Vinay Modi http://in.pyco n.org/.../.../ Sematicweb_ Python http://www. voicepitara.com
http://.../deleg ate/vinaymodi
hasProfile
hasTutorial
RDFLib graph merging
g.parse(inPyconSlides_rdf, format='n3') g1 = Graph() myns = Namespace('http://example.com/') # object of the triple in g1 is subject of a triple in g. g1.add(('http://vinaymodi.googlepages.com/', myns['hasTutorial'], inPyconSlides['Semanticweb_Python.pdf']) mgraph = g + g1
g1
RDFLib some possible things you can do
Creating named graphs Quoted graphs Fetching remote graphs and querying over them RDF Literals are XML Schema datatype; Convert Python datatype to RDF Literal and vice versa. Persistent datastore in MySQL, Sqlite, Redland, Sleepycat, ZODB, SQLObject Graph serialization in RDF/XML, N3, NT, Turtle, TriX, RDFa
End of the Tutorial
Thank you for listening patiently.
Contact: Vinay Modi Voice Pitara Technologies (P) Ltd [email protected]
(Queries for project development, consultancy, workshops, tutorials in Knowledge representation and Semantic Web are welcome)
