1. Database Management Systems
Faculty of Computer Science
Technion – Israel Institute of Technology
Database Management Systems
Course
Lecture 9:
RDF and SPARQL

2. The Semantic Web
“The Semantic Web is an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation.”
Tim Berners-Lee, James Hendler and Ora Lassila,
Scientific American, May 2001
“The Web was designed as an information space, with the goal that it should be useful not only for human-human communication, but also that machines would be able to participate and help. One of the major obstacles to this has been the fact that most information on the Web is designed for human consumption, and even if it was derived from a database with well defined meanings (in at least some terms) for its columns, that the structure of the data is not evident to a robot browsing the web. Leaving aside the artificial intelligence problem of training machines to behave like people, the Semantic Web approach instead develops languages for expressing information in a machine processable form.”
Tim Berners-Lee, Semantic Web Road map, 1998

3. More Technically ...
Vision: Web data will entail semantics in a manner that is understood (and processed, and linked) automatically by computers
The “Semantic Web” is the technical infrastructure
Unambiguous names for resources that may also bind data to real world objects (URI)
A common data model to describe/link resources (RDF)
Expressive language to access to data (SPARQL)
Languages for defining common vocabularies and rules for automated reasoning (e.g., politician is person) (RDFS, OWL)
Data providers should collaborate: properly publish their data and link it to existing data

4. RDF and SPARQL Recommendations
“Resource Description Framework”
1997: first public draft; 1999: recommendation (revised 2004); 2014: RDF1.1
SPARQL
Query language for RDF
“SPARQL Protocol and RDF Query Language”
Yes, recursive definition
2004: first public draft; 2008: recommendation; 2013: SPARQL1.1

5. Some Freely Available RDF Repositories
DBPedia (~1.2b triples)
“a crowd-sourced community effort to extract structured information from Wikipedia and make this information available on the Web”
Freebase (~340m triples)
“A community-curated DB of well-known people, places, and things”
DBLP (~58m triples)
Computer science bibliography
WordNet (~3m triples)
English lexical db: synonyms, antonyms, POS, ...
GeoNames (~94m triples)
“Covers all countries, contains over eight million placenames”
Yago (~120m triples)
Information from Wikipedia, Wordnet, Geonames

6. There’s More! Linked!

8. RDF Example from DBPedia
13253
1949

9. In XML Syntax
13253
<?xml version="1.0" encoding="utf-8" ?>
<rdf:RDF
xmlns:rdf="
xmlns:rdfs="
xmlns:dbp="
xmlns:dbo="
<rdf:Description rdf:about="
<dbp:students rdf:datatype="
<dbp:president rdf:resource=" />
</rdf:Description>
</rdf:RDF>

10. Querying (SPARQL) SELECT ?inst { ?inst dbp:president dbr:Peretz_Lavie
13253
PREFIX dbp:<
PREFIX dbr:<
SELECT ?inst {
?inst dbp:president dbr:Peretz_Lavie }
inst

11. The owl:sameAs Property
url
SELECT DISTINCT ?city {
?c dbo:type dbr:City.
?c dbp:name ?cn.
FILTER ( str(?cn)="Haifa").
?c owl:sameAs ?city. }
These URIs can be used for getting Haifa information from other RDF resources!

12. RDF Components URI vocabulary Description of directed graphs
Uniquely identify an object / type / concept
Basically, a string that can encode (hierarchies of) concepts in domains
Description of directed graphs
With URIs for node and edge labels
Edge = triple
Schema definition
Extend the vocabulary with reasoning over inheritance, domain/range of properties, …

13. Additional Quotes from Scientific A. 05/01
Two important technologies for developing the Semantic Web are already in place: eXtensible Markup Language (XML) and the Resource Description Framework (RDF).
XML lets everyone create their own tags ...
Meaning is expressed by RDF, which encodes it in sets of triples, each triple being rather like the subject, verb and object of an elementary sentence.
... These triples can be written using XML tags.
RDF [...] makes assertions that particular things (people, Web pages or whatever) have properties (such as "is a sister of," "is the author of") with certain values (another person, another Web page).

14. Ontologies Originally, a philosophical discipline
Science of Being (Aristotle, Metaphysics, IV, 1)
What characterizes being? Eventually, what is being?
An ontology is a formal specification of a shared conceptualization (T. R. Gruber)
Practically speaking, in the world of Semantic Web:
A public collection of entity and relationship names
An agreed-upon meaning (usually based on textual descriptions and context, e.g., HTML table or furniture table)
Identified by a known namespace URI
Rules on behavior (e.g., FOL)

15. Public Ontologies
Best practice is to define data using existing ontologies
Famous examples:
Dublin Core (media, Web resources)
FOAF: Friend of a Friend (people/social)
The Music Ontology
Good Relations (e-commerce)
Used by Best Buy for their RDF catalog

16. Main Lecture Resources
Lecture notes by Marcelo Arenas
RDF y SPARQL: Dos componentes básicos para la Web de datos
Lecture notes by Yaron Kanza
The Semantic Web

18. URI Uniform Resource Identifier Syntax has a specification in W3C
RFC 2396
A URL is a URI that points to a location on the network (e.g., server on the network)
e.g.,
A general URI needs not be an actual URL
e.g., urn:isbn:

19. Relevant Quotes from S. A. 05/2001
Because RDF uses URIs to encode this information in a document, the URIs ensure that concepts are not just words in a document but are tied to a unique definition that everyone can find on the Web.
For example, imagine that we have access to a variety of databases with information about people, including their addresses. If we want to find people living in a specific zip code, we need to know which fields in each database represent names and which represent zip codes.
RDF can specify that “(field 5 in database A) (is a field of type) (zip code),” using URIs rather than phrases for each term.

20. Example Revisited
13253
<?xml version="1.0" encoding="utf-8" ?>
<rdf:RDF
xmlns:rdf="
xmlns:rdfs="
xmlns:dbp="
xmlns:dbo="
<rdf:Description rdf:about="
<dbp:students rdf:datatype="
<dbp:president rdf:resource=" />
</rdf:Description>
</rdf:RDF>

21. Semantically, a URI always represents the expanded name
Terminology
<root
xmlns:h="
xmlns:f="
<h:table>
prefix h
local name
table
qualified name
h:table
namespace URI
expanded name
Semantically, a URI always represents the expanded name

22. Example Revisited http://dbpedia.org/ontology/country
13253
1949

23. Example Revisited (Clearer Now)
@prefix dbp: < .
@prefix dbr: < .
@prefix dbo: < .
dbo:country
dbr:Israel
dbr:Petah_Tikva
dbo:birthPlace
dbo:birthPlace
dbr:Technion_Israel_Institute_of_Technology
dbr:Peretz_Lavie
dbp:president
dbo:birthYear
dbp:students
1949
13253

25. RDF Graph An RDF graph is a set of triples
Example:
(dbr:technion, dbp:oresident, dbr:Lavie)
(dbr:technion, dbp:students, 13253)
Formally, we assume two sets:
U is a set of URIs, L is a set of literals
An RDF triple is a member of U×U×(U∪L)
An RDF graph is a finite set of RDF triples
predicate
subject
object U U U L

26. Running Example: RDF Graph
dbr:Nate_Huffman
dbp:team
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbo:height
2.159
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
2.05
URI
literal

27. Our Model is a Simplification
It does not represent blank nodes
These are nodes without URIs
It does not model datatypes
e.g., integer, float, string, date, ...
Further reading for interest: RDF books, online tutorials, W3C
dbr:Israel
dbo:country
dbo:address
dbr:Technion_Israel_Institute_of_Technology
dbo:zip
32000
dbo:city
dbr:Haifa

28. Representing RDF Graphs
There are several formats for representing RDF triples
We have seen examples of RDF/XML
Namespaces, Description, ...
JSON-LD
LD stands for Linked Data
N-Triple
Straightforward triples
Turtle
More compact than N-Triple

29. Representation Examples
<
< .
< dbpedia.org/property/students>"13253"^^< .
< < .
N-Triple
{ [
{ "
" :
[ { " ] ,
" : [ 870 ] ,
" : [ ] ] }
Turtle
<
dbp:country dbr:Israel ;
dbp:staff 870 ;
dbp:budget "US$381 ;
dbp:president dbr:Peretz_Lavie ;
dbp:undergrad ;
dbp:established ;
dbp:students ;
JSON-LD

31. Query Language for RDF? To query RDF, we can use existing languages
SQL (over the triple relation)
XQuery (over RDF/XML)
But:
We want QLs that are simpler and intuitive on graphs (triples)
e.g., graph patterns should be easily phrased
The open-world nature of the semantic Web makes incompleteness a first-class citizen
We should easily allow for optional information when exists (not NULL reasoning)

32. SPARQL Query Types We restrict to SELECT! SELECT ASK DESCRIBE
Returns mappings
ASK
Boolean – exists at least one mapping?
DESCRIBE
Includes a detailed description of found URIs
CONSTRUCT
Construct a new RDF via the query
We restrict to SELECT!
But the main idea is very similar on all

33. SPARQL Examples SELECT ?x ?n { ?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n.
?x dbo:height ?h.
FILTER(?h>2) }
SELECT ?x ?n {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n }
Maccabi players + nationalities
Restrict to players >2m
SELECT ?x ?n ?h ?t {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n
OPTIONAL {
?x dbp:team ?t.
?t rdf:type yago:NBA}
?x dbo:height ?h } }
SELECT ?x ?n ?h {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n
OPTIONAL {
?x dbo:height ?h } }
Add height optionally (left o. join)
Add both height and NBA memberships optionally

34. Try it on http://dbpedia.org/sparql
SELECT distinct ?x ?t ?h {
?x dbp:team <
OPTIONAL {
?x dbp:team ?t.
?t rdf:type yago:WikicatNationalBasketballAssociationTeams }
?x dbo:height ?h

36. RDF Query Model
Next, we will see a formal definition of the SPARQL syntax and semantics
Our definition covers the core of SPARQL, but omits many details
Does not cover blank nodes
Does not list the plethora of functions and Boolean predicates used for filtering
As usual, more details in books, online tutorials, W3C, etc.

37. Basic SPARQL Query SELECT ?x,?y,... { graph-pattern }
The semantics is based on the notion of a mapping
A mapping is a function μ that maps a set of variables, denoted vars(μ), to URIs and literals
μ : vars(μ) ⟶ U ∪ L
When applying a graph pattern P to an RDF graph G, the result is a set of mappings, denoted by P(G)
The result of the query is the restriction (projection) of these mappings to ?x,?y,...

38. Triple Patterns t(G) = { μ : vars(t) ⟶ U∪L | μ(t) ∈ G} S P O
URI / variable
URI / variable
URI / variable / literal
The simplest (atomic) pattern is a triple pattern
Semantics of a triple pattern t:
t(G) = { μ : vars(t) ⟶ U∪L | μ(t) ∈ G}
That is, t(G) contains all of the mappings μ such that:
vars(μ) consists of the variables in t, and
When replacing each variable ?x with the constant μ(?x), the triple t becomes a triple in G
In the definition we denote the new triple as μ(t)

39. Example 1 ?x dbp:team ?t dbr:Nate_Huffman dbp:team dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbo:height
2.159
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
2.05
?x dbp:team ?t x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
{ ?x ⟶ dbr:Nate_Huffman , ?t ⟶ dbr:Maccabi_TelAviv}
{ ?x ⟶ dbr:Nate_Huffman , ?t ⟶ dbr:Toronto_Raptors}
{ ?x ⟶ dbr:Gal_Mekel, ?t ⟶ dbr:Maccabi_...}
{ ?x⟶dbr:Gal_Mekel, ?t ⟶ dbr:Dallas_Mavericks}
{ ?x⟶dbr:Lior_Eliyahu, ?t ⟶ dbr:Maccabi_TelAviv}

40. Example 2 dbr:Nate_Huffman ?p ?o dbr:Nate_Huffman dbp:team
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbo:height
2.159
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
2.05
dbr:Nate_Huffman ?p ?o p o
dbp:team
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbo:height
2.159

41. ?x ∈ (vars(μ1) ∩ vars(μ2)) ⟹ μ1(?x)=μ2(?x)
Terminology
Two mappings μ1 and μ2 are compatible if they agree on their common values
?x ∈ (vars(μ1) ∩ vars(μ2)) ⟹ μ1(?x)=μ2(?x)
The (natural) join μ1⨝μ2 of two compatible mappings μ1 and μ2 is the mapping μ with:
vars(μ) = vars(μ1) ∪ vars(μ2)
μ(?x)=μi(?x) whenever ?x∈vars(μi)
The projection πX(μ) of a mapping μ to a set X of variables is the mapping λ with:
vars(λ) = vars(μ) ∩ X
λ(?x)=μ(?x) for all ?x in vars(λ)

42. Example: Compatibility and Join
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors x t
dbr:Nate_Huffman
dbr:Toronto_Raptors x t0
dbr:Nate_Huffman
dbr:Toronto_Raptors
compatible
incompatible x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv x0 t0
dbr:Gal_Mekel
dbr:Dallas_Mavericks
comp. x t x0 t0
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Gal_Mekel
dbr:Dallas_Mavericks
compatible x h
dbr:Nate_Huffman
2.159
incompatible x t h
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159 x0 t
dbr:Gal_Mekel
dbr:Dallas_Mavericks

43. Example: Projection μ π{x,t}(μ) { } x t dbr:Nate_Huffman
dbr:Toronto_Raptors x t
dbr:Nate_Huffman
dbr:Toronto_Raptors x t h
dbr:Nate_Huffman
dbr:Toronto_Raptors
2.159 x t
dbr:Nate_Huffman
dbr:Toronto_Raptors x0 t
dbr:Gal_Mekel
dbr:Dallas_Mavericks t
dbr:Dallas_Mavericks x0 t0
dbr:Gal_Mekel
dbr:Dallas_Mavericks
{ }

44. Equivalent Presentations of Mapping Setsx t h
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
dbr:Lior_Eliyahu
2.05
Collection of complete tables x t a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
yago:NBA
dbr:Toronto_Raptors
yago:EL
dbr:Lior_Eliyahu
The same! x t h a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
yago:NBA
yago:EL
dbr:Lior_Eliyahu
2.05
Missing Values

45. Algebraic Operations Ω1-Ω2 = {μ1∈Ω1 | no μ2∈Ω2 is compatible with μ1}
Let Ω1 and Ω2 be two sets of matches
Join: Ω1⨝Ω2 = {μ1⨝μ2 | μi∈Ωi ⋀ compatible(μ1,μ2)}
Difference:
Ω1-Ω2 = {μ1∈Ω1 | no μ2∈Ω2 is compatible with μ1}
Left outer join: Ω1⟕Ω2 = (Ω1⨝Ω2) ∪ (Ω1-Ω2)
Projection: πX(Ω1) = {πX(μ1) | μ1∈Ω1}
Filter (selection) for a Boolean condition F over matches
σF(Ω1) = {μ1∈Ω1 | μ1 satisfies F}

46. Example: Join ⨝ = x t dbr:Nate_Huffman dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu x h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 ⨝ = x t h
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
dbr:Lior_Eliyahu
2.05

47. Example: 2xJoin ⨝ ⨝ = x t dbr:Nate_Huffman dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu x h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 ⨝ t a
dbr:Toronto_Raptors
yago:NBA
dbr:Maccabi_TelAviv
yago:EL ⨝ x t h a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
yago:EL
dbr:Toronto_Raptors
yago:NBA
dbr:Lior_Eliyahu
2.05 =

48. Example: Difference (1)t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu x h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 - = x t
dbr:Gal_Mekel
dbr:Dallas_Mavericks

49. Example: Difference (2)h
dbr:Lior_Eliyahu
2.05 x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu - = t a
dbr:Dallas_Mavericks
yago:NBA x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors

50. Example: Difference (3)t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu a
yago:NBA
yago:EL
{ } - =

51. Example: Left Outer Join (1)
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu x h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 ⟕ = x t h
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
2.05

52. Example: Left Outer Join (2)
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
dbr:Utah_Jazz t a
dbr:Utah_Jazz
yago:NBA
dbr:Maccabi_TelAviv
yago:EL ⟕ = x t a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
yago:EL
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu
dbr:Utah_Jazz
yago:NBA

53. Example: Join, Difference, Outer Join (1)h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu ⨝ = t a
dbr:Toronto_Raptors
yago:NBA
dbr:Maccabi_TelAviv
yago:EL x t h a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
yago:NBA
yago:EL
dbr:Lior_Eliyahu
2.05

54. Example: Join, Difference, Outer Join (2)h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu - = t a
dbr:Toronto_Raptors
yago:NBA
dbr:Maccabi_TelAviv
yago:EL x t
dbr:Gal_Mekel
dbr:Dallas_Mavericks

55. Example: Join, Difference, Outer Join (3)h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05 x t
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
dbr:Toronto_Raptors
dbr:Gal_Mekel
dbr:Dallas_Mavericks
dbr:Lior_Eliyahu ⟕ = t a
dbr:Toronto_Raptors
yago:NBA
dbr:Maccabi_TelAviv
yago:EL x t h a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
yago:NBA
yago:EL
dbr:Lior_Eliyahu
2.05
dbr:Gal_Mekel
dbr:Dallas_Mavericks

56. Projection Example π?x,?h = x t h a dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
dbr:Toronto_Raptors
yago:NBA
yago:EL
dbr:Lior_Eliyahu
2.05
dbr:Gal_Mekel
dbr:Dallas_Mavericks
π?x,?h = x h
dbr:Nate_Huffman
2.159
dbr:Lior_Eliyahu
2.05
dbr:Gal_Mekel

57. Filter Example σ?a=yago:EL = x t h a dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
yago:EL
dbr:Toronto_Raptors
yago:NBA
dbr:Lior_Eliyahu
2.05
dbr:Gal_Mekel
dbr:Dallas_Mavericks
σ?a=yago:EL = x t h a
dbr:Nate_Huffman
dbr:Maccabi_TelAviv
2.159
yago:EL
dbr:Lior_Eliyahu
2.05
dbr:Gal_Mekel

58. SPARQL Filter Language
Atomic binary predicates (+scalar functions)
bound(?x)
Satisfied by μ if ?x is in vars(μ)
?x=?y, ?x=“Technion”
Satisfied by μ if involved variables are bound and equality holds
regex(str(?mbox),
Composition via Boolean operators
and, or, not

59. SPARQL Basic Syntax Pattern P Semantics: P(G) triple pattern t t(G)
P1 . P2
P1(G) ⨝ P2(G)
{P1} MINUS {P2}
P1(G) - P2(G)
P1 OPTIONAL {P2}
P1(G) ⟕ P2(G)
{P1} UNION {P2}
P1(G) ∪ P2(G)
{P1 FILTER ( F )}
σFP1(G)
SELECT ?x1,...,?xk WHERE {P1}
π{x1,...,xk}(P1(G))
optional

60. Examples Revisited SELECT ?x ?n { ?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n.
?x dbo:height ?h.
FILTER(?h>2) }
SELECT ?x ?n {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n }
triple, join, projection
triple, join, projection, filter
SELECT ?x ?n ?h ?t {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n
OPTIONAL {
?x dbp:team ?t.
?t rdf:type yago:NBA}
?x dbo:height ?h } }
SELECT ?x ?n ?h {
?x dbp:team dbr:Maccabi_TelAviv.
?x dbp:nationality ?n
OPTIONAL {
?x dbo:height ?h } }
triple, join, projection, l.o.join
triple, join, projection, l.o.join

61. Ex. 1
dbr:Abraham_Lincoln
dbo:party
dbr:Republican_Party
dbr:Whig_Party
dbo:profession
dbr:Lawyer
rdf:type
yago:PresidentsOfTheUnitedStates
dbr:Franklin_D._Roosevelt
dbr:Democratic_Party
dbr:Richard_Nixon
dbr:Bill_Clinton
SELECT ?president {
?president rdf:type yago:PresidentsOfTheUnitedStates.
?president dbo:profession dbr:Lawyer. }
Lincoln
Roosevelt
Nixon
Clinton ?

62. Ex. 2
dbr:Abraham_Lincoln
dbo:party
dbr:Republican_Party
dbr:Whig_Party
dbo:profession
dbr:Lawyer
rdf:type
yago:PresidentsOfTheUnitedStates
dbr:Franklin_D._Roosevelt
dbr:Democratic_Party
dbr:Richard_Nixon
dbr:Bill_Clinton
SELECT ?president {
?president rdf:type yago:PresidentsOfTheUnitedStates.
?president dbo:profession dbr:Lawyer.
OPTIONAL { ?president dbo:party ?party } }
Lincoln
Roosevelt
Nixon
Clinton ?

63. Ex. 3
dbr:Abraham_Lincoln
dbo:party
dbr:Republican_Party
dbr:Whig_Party
dbo:profession
dbr:Lawyer
rdf:type
yago:PresidentsOfTheUnitedStates
dbr:Franklin_D._Roosevelt
dbr:Democratic_Party
dbr:Richard_Nixon
dbr:Bill_Clinton
SELECT ?president { {
?president rdf:type yago:PresidentsOfTheUnitedStates.
?president dbo:profession dbr:Lawyer.
OPTIONAL { ?president dbo:party ?party } }
MINUS {dbr:Richard_Nixon dbo:party ?party}
Lincoln
Roosevelt
Nixon
Clinton ?

64. Ex. 4 Write SPARQL queries that:
dbr:Abraham_Lincoln
dbo:party
dbr:Republican_Party
dbr:Whig_Party
dbo:profession
dbr:Lawyer
rdf:type
yago:PresidentsOfTheUnitedStates
dbr:Franklin_D._Roosevelt
dbr:Democratic_Party
dbr:Richard_Nixon
dbr:Bill_Clinton
Write SPARQL queries that:
Find the presidents who were in one or more of Lincoln’s parties
Find the presidents who were in none of Lincoln’s parties
For each president in (1) and (2), add his dbp:spouse if exists

66. From Graphs to Datasets
An important ability in SPARQL is the ability to query multiple graphs (from multiple sources)
This is formalized as follows
An RDF dataset D is a pair (G0,Ext), where:
G0 is the default RDF graph
Ext is a finite mapping of URIs to RDF graphs
These are the other graphs that can be referenced by means of URIs G0
Ext(u1)
Ext(u2)
Ext(u3)

67. SPQRQL over RDF Datasets
Let D=(G0,Ext) be an RDF dataset, G be an RDF graph (G=G0 or G∈Ext), and P be a graph pattern
If u be a URI in the domain of Ext, then now Q={GRAPH u {P}} is also a graph pattern
Q(G) = P(Ext(u))
Q={GRAPH ?x {P}} is now also a graph pattern
Q(G) = ∪u∈dom(Ext)({?x⟶u} ⨝ P(Ext(u)))
We can define the eval of a pattern P' over D:
P'(D) = P'(G0)
Ignore G!

68. Adding Graph to Basic Syntax
Pattern P
Semantics: P(G)
triple pattern t
t(G)
P1 . P2
P1(G) ⨝ P2(G)
{P1} MINUS {P2}
P1(G) - P2(G)
P1 OPTIONAL {P2}
P1(G) ⟕ P2(G)
{P1} UNION {P2}
P1(G) ∪ P2(G)
{P1 FILTER ( F )}
σFP1(G)
SELECT ?x1,...,?xk {P1}
π{x1,...,xk}(P1(G))
GRAPH u {P1}
P1(Ext(u))
GRAPH ?x {P1}
∪u∈dom(Ext)({?x⟶u} ⨝ P1(Ext(u)))

69. Example Default (i.e., G0) dbr:Ashdod dbp:mayor dbr:Yehiel_Lasri
owl:sameAs
ex:Yehiel_Lasri
dbr:Haifa
dbr:Yona_Yahav
ex:Yona_Yahav
ex:Yona_Yahav
ex:isAffiliatedTo
ex:Kadima
ex:Yona_Yahav
ex:isAffiliatedTo
ex:Labor
SELECT ?m ?g ?party
FROM NAMED <
FROM NAMED < {
?city dbp:mayor ?m1
?m1 owl:sameAs ?m
OPTIONAL {
GRAPH ?g {?m ex:isAffiliatedTo ?party} }} m g
party
ex:Yehiel_Lasri
ex:Yona_Yahav
ex:Kadima
ex:Labor
Ext

70. Mammals from dbpedia.org
Mammals in Hebrew:
SELECT distinct ?x {
?x dbp:classis dbr:Mammal. }
SELECT distinct ?x, ?h {
?x dbp:classis dbr:Mammal.
?x rdfs:label ?h.
FILTER(LANG(?h) = "he") } x … x h

71. Mammals from dbpedia.org + Linked
SELECT distinct ?x, ?h {
?x dbp:classis dbr:Mammal.
?x owl:sameAs ?y.
GRAPH ?g {
?y rdfs:label ?h .
FILTER(LANG(?h) = "he") } x h
"אנומלור
"גחן
"סלנויה חומת
"מושק
"ארמדיל חשוף-זנב
"איילון
"ויסקאשה

72. Beautifying SELECT distinct str(?e) as ?eng, str(?h) as ?heb {
?x dbp:classis dbr:Mammal.
?x owl:sameAs ?y.
?x rdfs:label ?e.
GRAPH ?g {
?y rdfs:label ?h. }
FILTER(LANG(?h)="he" && LANG(?e)="en")

74. Additional Features (SPARQL 1.1)
Aggregate Queries
Property Path
SELECT ?city, count(?x) AS ?c {
?x rdf:type dbo:Scientist.
?x dbo:birthPlace ?city.
?city dbo:country dbr:Israel. }
GROUP BY ?city
ORDER BY DESC(count(*))
select ?s {
?s rdf:type dbo:Scientist.
?s (dbo:doctoralAdvisor)+ dbr:Alonzo_Church. } s
dbr:Alan_Turing
dbr:Andrej_Bauer
dbr:Angus_Macintyre
dbr:Dana_Scott
dbr:Dov_Gabbay
dbr:J._Barkley_Rosser
dbr:Jack_Copeland
dbr:Jeff_Paris
dbr:Kurt_Mehlhorn
dbr:Martin_Davis
dbr:Martin_Hyland s
dbr:Michael_Fourman
dbr:Michael_O._Rabin
dbr:Paul_Ernest
dbr:Peter_Mosses
dbr:Robert_Lee_Constable
dbr:Robin_Gandy
dbr:Saharon_Shelah
dbr:Stephen_Cole_Kleene
dbr:Susanne_Albers
dbr:William_Boone
dbr:Yiannis_N._Moschovakis
city c
dbr:Tel_Aviv 30
dbr:Haifa 16
dbr:Rehovot 6
dbr: Petah_Tikva 4
dbr:Tiberias 2
dbr:Safed 1
dbr:Nazareth

75. Advisory Graph dbr:Alonzo_Church dbo:doctoralAdvisor dbr:Alan_Turing
dbr:Robin_Gandy
dbr:Jeff_Paris
dbr:Michael_O._Rabin
dbo:doctoralAdvisor
dbr:Saharon_Shelah
dbr:Dov_Gabbay

77. RDF Schema RDF Schema (RDFS) enriches RDF with semantics
Types/subtypes, classes/subclasses
Properties and sub-properties
Typing of properties
A simple extension to describe ontologies
Very limited compared to other concepts
Monotone – does not say what “cannot” happen
Very different notion from SQL schema, DTD, ...
Not outside RDF, extends RDF
Why is it useful?
General reasoning/integration power
Allows to automatically infer implicit information

78. Example (Wikipedia) SELECT ?x { ?x rdf:type ex:animal } ex:dog1
ex:cat1
ex:cat
ex:zoo1
zoo:host
ex:cat2
SELECT ?x {
?x rdf:type ex:animal } x
ex:dog1
rdf:type
ex:dog1
ex:animal
rdf:type
ex:cat1
ex:cat
ex:cat2
ex:zoo1
zoo:host

79. Example (Wikipedia) SELECT ?x { ?x rdf:type ex:animal } ex:dog1
ex:cat1
ex:cat
ex:zoo1
zoo:host
ex:cat2
rdfs:subClassOf
rdfs:range
SELECT ?x {
?x rdf:type ex:animal } x
ex:dog1
ex:cat1
ex:cat2
rdf:type
ex:dog1
ex:animal
rdfs:subClassOf
rdfs:range
rdf:type
ex:cat1
ex:cat
ex:cat2
ex:cat
zoo:host

80. We Discuss Four RDFS Properties
p rdfs:domain C
Example: ex:teaches rdfs:domain ex:professor
p rdfs:range C
Example: ex:teaches rdfs:range ex:course
p rdfs:subPropertyOf q
Example: ex:motherOf rdfs:subPropertyOf ex:parentOf
C rdfs:subClassOf D
Example: ex:pizza rdfs:subClassOf ex:Food

81. Inference System for RDFS
Rules for inferring additional triples from existing RDF/RDFS triples:
Given
⇒ Infer
(p,rdfs:domain,C) , (s,p,o)
(s,rdf:type,C)
(p,rdfs:range,C) , (s,p,o)
(o,rdf:type,C)
(p,rdfs:subPropertyOf,q) , (s,p,o)
(s,q,o)
(C,rdfs:subClassOf,D) , (s,rdf:type,C)
(s,rdf:type,D)
Transitivity
(p,rdfs:subPropertyOf,q) , (q,rdfs:subPropertyOf,r)
(p,rdfs:subPropertyOf,r)
(C,rdfs:subClassOf,D) , (D,rdfs:subClassOf,E)
(C,rdfs:subClassOf,E)

82. Example dbr:Israel ex:minister ex:Gila_Gamliel ex:finance_minister
ex:Moshe_Kahlon
rdfs:subPropertyOf
rdfs:range
dbr:politician
rdfs:subClassOf
dbo:person
dbr:Israel
ex:minister
ex:Moshe_Kahlon
ex:Gila_Gamliel
drf:type
dbr:politician
ex:Gila_Gamliel
drf:type
dbo:person
ex:Moshe_Kahlon
drf:type
dbr:politician
ex:Moshe_Kahlon
drf:type
dbo:person
IL minister
IL finance_minister
politician
IL minister
person
politician
person

83. RDFS Closure
The RDFS closure of an RDF graph G is the graph obtained from G by iteratively adding triples inferred from the rules until impossible (fixpoint)
We denote this graph by cl(G)
A SPARQL engine that supports RDFS gives the result of cl(G)
PRDFS(G) = P(cl(G))

84. Example SELECT ?x { PRDFS(G) ?x rdf:type dbo:Scientist.
dbr:Moshe_Vardi
dbo:doctoralAdvisor
dbr:Catriel_Beeri
rdfs:domain
dbo:Scientist
dbo:birthPlace
dbr:Haifa
dbr:Ron_Kimmel
rdf:type
SELECT ?x {
?x rdf:type dbo:Scientist.
?x dbo:birthPlace dbr:Haifa }
PRDFS(G) x
dbr:Ron_Kimmel
dbr:Moshe_Vardi
P(G) x
dbr:Ron_Kimmel

85. RDFS Expressiveness Limitations
No negation
Cannot state disjointness (e.g., animals / people)
Limited containment relationship
For example, every mammal is either an animal or a person
No cardinality/inclusion constraints
Every mammal has exactly two biological parents
Every course has at least one lecturer
No logical constraints on relations
For example, inverse (parent/child), transitivity

86. Beyond RDFS: OWL
OWL (Web Ontology Language) is designed to provide rich expressiveness for ontology specification and reasoning
“OWL is now the most used KR language in the history of AI…” (Jim Hendler)
Several classes (different expressiveness / complexity tradeoffs)
OWL Full, OWL DL, OWL Lite
In all versions, specs are in RDF

87. Examples foaf:made owl:inverseOf foaf:maker
ex:friend rdf:type owl:SymmetricProperty
geo:located rdf:type owl:TransitiveProperty
Additional properties:
owl:FunctionalProperty, owl:ComplementOf, owl:IntersectionOf