[PDF]

Ontology Matching PDF Free Download

1 / 57
0 views•57 pages

Ontology Matching PDF Free Download

Ontology Matching PDF free Download. Think more deeply and widely.

Ontology Matching
!"#$%"&' ()$%*
++*,)-."*+/++
!"#$%&'()*)'+',%-)#*".'/%0'123245267
Goals of today’s lecture
•8#*9:&;.%'<=)*':#*:>:?@'-)*.="#?'"A')#&'<=@'"*'"A'"-B:9*)#*
•("A.;AA'B9).*".)>')#&'-%*=:&:>:?".)>'"AA;%A
•(%A.9"0%'&"CC%9%#*'*%.=#"D;%A'C:9':#*:>:?@'-)*.="#?
•EF)>;)*%'*=%'D;)>"*@':C':#*:>:?@'-)*.="#?
Introduction
What is an ontology?
•6#':#*:>:?@'*@B".)>>@'B9:F"&%A')'F:.)0;>)9@ &%A.9"0"#?')'&:-)"#':C'
"#*%9%A*')#&')'AB%."C".)*":#':C'*=%'-%)#"#?':C'*%9-A'"#'*=)*'
F:.)0;>)9@
•(%B%#&"#?':#'*=%'B9%."A":#':C'*="A'AB%."C".)*":#G'*=%'#:*":#':C'
:#*:>:?@'%#.:-B)AA%A'A%F%9)>'&)*)')#&'.:#.%B*;)>'-:&%>AG'
"#.>;&"#?G'A%*A':C'*%9-AG'.>)AA"C".)*":#AG'*=%A);9"G'&)*)0)A%'A.=%-)AG'
:9'C;>>@')H":-)*"I%&'*=%:9"%AJ'
Various forms of ontologies
Euzenat, J. & Shvaiko, P. (2013). Ontology matching (2nd ed.). Springer Berlin Heidelberg.
Ontologies are not Reality
•K#*:>:?"%A')9%')'.:#*%H*5&%B%#&%#*'B9:L%.*":#'1-:&%>7':C'*=%'M%)>"*@
•("CC%9%#*':#*:>:?"%A'-"?=*'-:&%>'*=%'A)-%'1A"-">)97':9'="?=>@'9%>)*%&'
&:-)"#AG'0;*'*=%@'-"?=*
•0#12#.3*4'11#-#$3*3)565*)$4*-#7&'-#8#$35*91*)::2'.)3'9$5
•;9229<*4'11#-#$3*.9$=#$3'9$5*)$4*-#53-'.3'9$5
Linked Open Data cloud
!""#$%&&'()*+'(,)-./"&
0( 1"21$2324/$$526,"232
4/3.1.78,'2(./9
The heterogeneity problem
•,@#*).N. =%*%9:?%#%"*@
•>5'$%*4'?#-#$3*9$3929%@*2)$%&)%#5*A#B%B*C,DE*F(DG
•O%9-"#:>:?".)> =%*%9:?%#%"*@
•H'?#-#$3*3#-85*-#1#-*39*3"#*5)8#*.9$.#:3
•!)8#*3#-8*4#5.-'I#5*4'?#-#$3*.9$.#:35
•P:#.%B*;)> 1,%-)#N.7'=%*%9:?%#%"*@
•H'?#-#$.#*'$*granularity, coverage, perspec2ve
•,%-":N. 1Q9)?-)N.7'=%*%9:?%#%"*@
•H'?#-#$3*'$3#-:-#3)J9$5*<-3 4'?#-#$3*.9$3#K3
Ontology matching to reduce
terminological &conceptual
heterogeneity
What is ontology matching?
•K#*:>:?@'-)*.="#?G'"A'*=%'B9:.%AA':C'&%*%9-"#"#?'.:99%AB:#&%#.%A'
0%*<%%# .:#.%B*A "# :#*:>:?"%AJ
Pambrix, Patrick & Ivanova, Valentina. (2013). A unified approach for debugging is-a structure and mappings in networked taxonomies. Journal of
biomedical semantics. 4. 10. 10.1186/2041-1480-4-10.
Correspondence
•R"F%#'*<:':#*:>:?"%A'𝑜)#&'𝑜’G')'.:99%AB:#&%#.% 0%*<%%#'𝑜)#&'𝑜’ "A'
&%C"#%&')A < 𝑖𝑑, 𝑒, 𝑒’, 𝑟, 𝑛 >G
<=%9%'
•𝑖𝑑 '5 )&$'7&# '4#$JL#- 91 3"# .9--#5:9$4#$.#
•𝑒)$4*𝑒’)-#*#$3'3'#5*91*𝑜)$4*𝑜’ -#5:#.3'=#2@E*#B%BE*.2)55#5E*'$53)$.#5*
•𝑟'5*)*-#2)3'9$E*#B%BE*#7&'=)2#$.#A=GE*89-#*%#$#-)2*A⊒)E*89-#*5:#.'1'.*A⊑)E*
4'5M9'$3$#55 A⊥)E*#3.B
•𝑛'5*)*.9$1'4#$.#*8#)5&-#*A3@:'.)22@*'$*)*-)$%#*91*N/EOPG*19-*3"#*
.9--#5:9$4#$.#*I#3<##$*𝑒)$4*𝑒’.
Correspondence relations
•K/!
•9<2Q#7&'=)2#$3R2)55E*9<2Q#7&'=)2#$3S-9:#-3@
•9<2Q5)8#T5E*9<2Q4'?#-#$3;-98E*9<2QT22H'?#-#$3
•,SK,
•5695Q#K).3,)3."E*5695Q.295#,)3."
•5695QI-9)4,)3."E*5695Q$)--9<,)3."
•5695Q-#2)3#4,)3."
Alignment
•R"F%#'*<:':#*:>:?"%A'𝑜)#&'𝑜’G')#')>"?#-%#* 0%*<%%#'𝑜)#&'𝑜’:
•'5*)*5#3*91*.9--#5:9$4#$.#5*9$*𝑜)$4*𝑜’
•<'3"*598#*)44'3'9$)2*8#3)4)3)*
•multiplicity: 1:1, 1:m, m:n
•method
•date
•creator
•…
Terminology
•T)*.="#? "A'*=%'B9:.%AA':C'C"#&"#?'9%>)*":#A="BA':9'.:99%AB:#&%#.%A'
0%*<%%#'%#*"*"%A':C'&"CC%9%#*':#*:>:?"%AJ'
•6>"?#-%#* "A')'A%*':C'.:99%AB:#&%#.%A'0%*<%%#'*<:':9'-:9%'1"#'.)A%'
:C'-;>*"B>%'-)*.="#?7':#*:>:?"%AJ'O=%')>"?#-%#*'"A'*=%':;*B;*':C'*=%'
-)*.="#?'B9:.%AAJ'
•P:99%AB:#&%#.% "A'*=%'9%>)*":#'A;BB:A%&'*:'=:>&')..:9&"#?'*:')'
B)9*".;>)9'-)*.="#?')>?:9"*=-':9'"#&"F"&;)>G'0%*<%%#'%#*"*"%A':C'
&"CC%9%#*':#*:>:?"%AJ'
•T)BB"#? "A'*=%':9"%#*%&G':9'&"9%.*%&G'F%9A":#':C')#')>"?#-%#*J'
Ontology matching applications
http://tutorial.ontologymatching.org/
Data Integration
:3"320(,;+/2<
:3"320(,;+/2=
:3"320(,;+/2>2
:3"32?."/7;3"1(. @.181/)2:3"32
•Title%2A3;;B2C(""/;23.)2"!/2C!1'($(#!/;D$20"(./
•Author%2E-2F-2G(H'1.7
•Subjects: E,I/.1'/281+"1(.5283."3$B281+"1(.
•Dewey%2J=>-K<L M* N.7'1$!281+"1(.526/"H//.2<KLO23.)2<KKK
•Publication details%2P(.)(.2%2Q'((4$6,;B52<KKR2=SSL2#;1."1.7-
•ISBN%2SRLRORLLR=2T#6UV
Let’s look at an example
•Title%2A3;;B2C(""/;23.)2"!/2C!1'($(#!/;D$20"(./
•Author%2E-2F-2G(H'1.7
•Illustrator%2W!(43$2W3B'(;2
•Publication year%2<KKR
•Publisher%2Q'((4$6,;B2T@FV
•ISBN%2S*RLRO*>=XK*K
•Title%2A3;;B2C(Y/;23.)2"!/2$(;+/;/;D$2$"(./
•Uniform ?tle%2A3;;B2C(Y/;23.)2"!/2#!1'($(#!/;D$2$"(./
•Personal name%2G(H'1.752E-2F-523,"!(;
•Published/Produced%2Z/H2[(;U52Z[2%20+!('3$\+2?.+-52]=S<J^
•ISBN%2KRJ<>>J=KK<LL2T#3#/;63+UV52<>>J=KK<L_2T#3#/;63+UV
•LC Subjects%2`1a3;)$bc1+\(.52d371+*ec1+\(.520+!(('$*c1+\(.
•Dewey class no.: J=>&-K<L2]c1+^
•Summary%2G/$+,/)28;(42"!/2(,";37/(,$2./7'/+"2(82!1$23,."23.)2
,.+'/5232B(,.726(B2H1"!2327;/3"2)/$\.B2#;(I/$2!1$2H(;"!2H!1'/2
3Y/.)1.72A(7H3;"$20+!(('28(;2`1"+!+;3f23.)2`1a3;);B-
•Title%2A3;;B2C(""/;23.)2"!/2C!1'($(#!/;D$20"(./
•Other title%2A3;;B2C(""/;23.)2"!/2$(;+/;/;D$2$"(./
•Author%2E-2F-2G(H'1.7
•Illustrator%2W!(43$2W3B'(;2
•Publication year%2<KKR
•Publisher%2Q'((4$6,;B2T@FV
•ISBN%2S*RLRO*>=XK*K52SRLRORLLR=2T#6UV52
KRJ<>>J=KK<LL2T#3#/;63+UV52<>>J=KK<L_2T#3#/;63+UV
•Subjects: E,I/.1'/281+"1(.5283."3$B281+"1(.52`1a3;)$b
c1+"1(.52d371+*ec1+"1(.520+!(('$*c1+"1(.
•Dewey%2J=>-K<L
•Summary%2G/$+,/)28;(42"!/2(,";37/(,$2./7'/+"2(82!1$2
3,."23.)2,.+'/5232B(,.726(B2H1"!2327;/3"2)/$"1.B2#;(I/$2
!1$2H(;"!2H!1'/23""/.)1.72A(7H3;"$20+!(('28(;2
`1"+!+;38"23.)2`1a3;);B-
After data integration
Data Integration
•,.=%-)'-)*.="#?'C:.;A%A':#'C"#&"#?'*=%'.:99%AB:#&%#.%')-:#?'
A.=%-)'%>%-%#*A'"#'*<:'A%-)#*".)>>@'.:99%>)*%&'A.=%-)*)
•,.=%-)'-)BB"#?'&%A.9"0%A'=:<')'A:;9.%'&)*)0)A%'A.=%-)'9%>)*%A'*:'
)'*)9?%*'&)*)0)A%'A.=%-)
•M%.:9&'>"#$)?%'1)>A:'$#:<#')A'%#*"*@'9%A:>;*":#')#&'&%&;B>".)*":#7'
"&%#*"C"%A'9%.:9&A'*=)*'9%C%9'*:'*=%'A)-%'>:?".)>'%#*"*@
•()*)'C;A":#'C:.;A%A':#'9%A:>F"#?'.:#C>".*A')#&'&%*%9-"#"#?'*=%'*9;%'
&)*)'F)>;%AG'>%F%9)?"#?'"#C:9-)*":#'"#'=%*%9:?%#%:;A'&)*)'A:;9.%A
X. Wang, L.M. Haas and A. Meliou, Explaining Data Integration, IEEE Data Eng. Bull. 41(2) (2018), 47–58
Data integration challenges
•Attributes may have different names-> entity attributes need to be mapped
Wikipedia
British Library
Library of Congress
•
W1"'/
•
g,"!(;
•
?'',$";3"(;
•
C,6'1+3"1(.2B/3;
•
C,6'1$!/;
•
?0QZ
•
W1"'/
•
g,"!(;
•
0,6h/+"$
•
:/H/B
•
C,6'1+3"1(.2)/"31'$
•
?0QZ
•
W1"'/
•
@.18(;42"1"'/
•
C/;$(.3'2.34/
•
C,6'1$!/)&C;(),+/)
•
Pi2$,6h/+"$
•
?0QZ
•
:/H/B2+'3$$2.(-
•
0,443;B
Data integration challenges
•6**9"0;*%'F)>;%A')9%'#:*'*=%'A)-%'5U'"#.:#A"A*%#.@'#%%&A'*:'0%'9%A:>F%&'
&;9"#?'merging
Wikipedia
British Library
Library of Congress
W1"'/2
A3;;B2C(""/;23.)2"!/2
C!1'($(#!/;D$20"(./
A3;;B2C(""/;23.)2"!/2
C!1'($(#!/;D$20"(./
A3;;B2C(""/;23.)2"!/2$(;+/;/;D$2
$"(./
?0QZ
S
*RLRO*>=XK*K
SRLRORLLR=2T#6UV
KRJ<>>J=KK<LL2T#3#/;63+UV52
<>>J=KK<L_2T#3#/;63+UV
0,6h/+"
E,I/.1'/281+"1(.5283."3$B2
81+"1(.
`1a3;)$
bc1+"1(.52d371+*e
c1+"1(.52
0+!(('$
*c1+"1(.
:/H/B
J=>-K<L
J=>&-K<L2]c1+^
C,6'1$!1.72
$"3",$
•
C,6'1+3"1(.2B/3;%2<KKR
•
C,6'1$!/;%2Q'((4$6,;B
T@FV
•
C,6'1+3"1(.2)/"31'$%2
P(.)(.2%2Q'((4$6,;B52
<KKR2=SSL2#;1."1.7-
•
C,6'1$!/)&C;(),+/)%2Z/H2
[(;U52Z[2%20+!('3$"1+2?.+-52
]=S<J^
Data integration challenges
•,*9;.*;9)>'=%*%9:?%#%"*@
Castano S., Ferrara A., Montanelli S., Varese G. (2011) Ontology and Instance Matching. https://doi.org/10.1007/978-3-642-20795-2_7
Data integration challenges
•T;>*"B>"."*@
•OQOE*OQ$E*8QOE*9-*8Q$U
•V"'5*)::2'#5*39*I93"*#$3'3'#5*)$4*)33-'I&3#5
Euzenat, J. & Shvaiko, P. (2013). Ontology matching (2nd ed.). Springer Berlin Heidelberg.
Data integration challenges
•T)*.="#?'A.:9%')#&'*=9%A=:>&
•W<'6'Q3'32#X*WI2Q3'32#X*OB/
•WI2QS&I2'.)3'9$Y4#3)'25X*W29.QS&I2'5"#4YS-94&.#4X*/BZ
•;)25#*:95'3'=#*=5*1)25#*$#%)3'=#
•()*)'A%*A')9%'*::'0"?'5U',.)>)0">"*@'"AA;%A
•D'I-)-@*91*R9$%-#55*")5*O[,\*I996*-#.9-45E*<"'2#*]-'3'5"*D'I-)-@*")5*+^,\*
)$4*(9-24R)3 _//,\`
•F:3'8'5)3'9$5
•Reduction of the number of comparisons
•Reduction of the cost of each comparison
Schema and instance matching
Ontology matching
•6'A.=%-) "A')'A*9;.*;9%':C'-%*)&)*) &%A.9"0"#?'=:<'&)*)G'"J%JG'
"#A*)#.%AG'.)#'0%'A*:9%&G')..%AA%&G')#&'"#*%9B9%*%&'0@';A%9A')#&'
)BB>".)*":#AJ'
•,.=%-)'-)*.="#?
•("#3"#-*3<9*.9$.#:35*A#B%BE*book =5*manuscriptG*9-*3<9*:-9:#-3'#5*A#B%B*
birthplace =5*hometownG*)-#*3"#*5)8#
•8#A*)#.%'-)*.="#?
•("#3"#-*4'11#-#$3*'$53)$.#5*-#1#-*39*3"#*5)8#*-#)2a<9-24*#$3'3@*'$*)*%'=#$*
498)'$*A#B%BE*)*:#-59$E*)*:2).#E*)*89='#E*)*I996E*#3.BG
Classification of schema matching techniques
Euzenat, J. & Shvaiko, P. (2013). Ontology matching (2nd ed.). Springer Berlin Heidelberg.
Categories of matching techniques
•E>%-%#*5>%F%> *%.=#"D;%A'.:#A"&%9':#*:>:?@'%#*"*"%A':9'*=%"9'"#A*)#.%A'
"#'"A:>)*":#'C9:-'*=%"9'9%>)*":#A'<"*=':*=%9'%#*"*"%A':9'*=%"9'"#A*)#.%A'
•,*9;.*;9%5>%F%> *%.=#"D;%A'.:#A"&%9'*=%':#*:>:?@'%#*"*"%A':9'*=%"9'
"#A*)#.%A'*:'.:-B)9%'*=%"9'9%>)*":#A'<"*=':*=%9'%#*"*"%A':9'*=%"9'
"#A*)#.%AJ'
Element-level techniques
•,*9"#?50)A%&V'-)*.="#?'#)-%A':9'&%A.9"B*":#A':C'%#*"*"%A
•!"#?;"A*".50)A%&V';A%'W!QG'>%H".:#AG':9'&:-)"#'AB%."C".'*=%A);9"'*:'
-)*.='<:9&A'0)A%&':#'>"#?;"A*".'9%>)*":#A'1=:-:#@-@G'A@#:#@-@G'
B)9*:#:-@G'%*.7':9'%HB>:"*"#?'-:9B=:>:?".)>'B9:B%9*"%A
•P:#A*9)"#%&50)A%&V'*)$%'"#*:')..:;#*'"#*%9#)>'.:#A*9)"#*A')BB>"%&'*:'
*=%'&%C"#"*":#A':C'%#*"*"%AG'%J?J'*@B%AG'.)9&"#)>"*@':C'B9:B%9*"%AG'%*.J'
•M%A:;9.%50)A%&V';A%'%H*%9#)>';BB%95>%F%>':9'&:-)"#5AB%."C".'
:#*:>:?"%A'*:'09"&?%'.>)AA%A'*:'0%'-)*.=%&J'
Element-level techniques: string-based
•!%F%#A=*%"# &"A*)#.%'9%B9%A%#*A'*=%'#;-0%9':C'"#A%9*":#AG'&%>%*":#AG')#&'
A;0A*"*;*":#A'9%D;"9%&'*:'.=)#?%':#%'<:9&'*:')#:*=%9J
•;9-*#K)8:2#E*3"#*D#=#$5"3#'$ 4'53)$.#*I#3<##$*b6'33#$b*)$4*b5'33'$%b*'5*cQ*
•kitten → sitten (substitution of "s" for "k"),
•sitten→ sittin (substitution of "i" for "e"),
•sittin → sitting(insertion of "g" at the end).
•()-%9);5!%F%#A=*%"# &"A*)#.%'.:;#*A'*9)#AB:A"*":#A')A')'A"#?>%'%&"*
•;9-*#K)8:2#E*4)8#-)&Y2#=#$5"3#'$Y4'53)$.#Ad1'5"dE*d'15"dG*ee*O*<"'2#*
2#=#$5"3#'$Y4'53)$.#Ad1'5"dE*d'15"dG*ee*+
Element-level techniques: string-based
•X)--"#?'&"A*)#.%'0%*<%%#'*<:'%D;)>5>%#?*='A*9"#?A':C'A@-0:>A'"A'*=%'
#;-0%9':C'B:A"*":#A')*'<=".='*=%'.:99%AB:#&"#?'A@-0:>A')9%'&"CC%9%#*J
•;9-*#K)8:2#E*3"#*f)88'$%*4'53)$.#*I#3<##$*b6)-92$b*)$4*b6#-53$b*'5*cB
•W5?9)-
•V)6#5*)5*'$:&3*3<9*53-'$%5*)$4*.)2.&2)3#*3"#*$&8I#-*91*3"#*.9889$*$a%-)85*
I#3<##$*3"#8E*$9-8)2'5#4*I@*8)KA2#$%3"A53-'$%OGE*2#$%3"A53-'$%+GG
•V-'%-)85*19-*8)3."'$% '5*8)3E*)3.E*3."E*."'E*"'$E*'$%
Element-level techniques: linguistic-based
•O:$%#"I)*":#
•S)-5#5*$)8#5*'$39*396#$5*I@*-#.9%$'g'$%*:&$.3&)3'9$E*.)5#5
•53-'$%aI)5#4*8#3"945*aX*N53-'$%E*I)5#4E*8#3"945P
•!%--)*"A)*":#
•T$)2@5#5*89-:"929%'.)22@*396#$5*39*1'$4*3"#'-*I)5'.*19-85
•8#3"945 aX*8#3"94
•M%-:F%'A*:B'<:9&A
•)E*"#E*3"#8E*I@E*1-98
Element-level techniques: phonetic-based
•Q=:#%*".'%#.:&"#?
•V"#*%9)2*'5*19- "989:"9$#5 39*I# #$.94#4 39*3"#*5)8#*-#:-#5#$3)3'9$*59*
3")3*3"#@*.)$*I#*8)3."#4*4#5:'3#*8'$9-*4'11#-#$.#5*'$ 5:#22'$%B
•!9&$4#K '5*) :"9$#3'.*)2%9-'3"8 19- '$4#K'$% $)8#5*I@*59&$4E*
)5 :-9$9&$.#4 '$*h$%2'5"B
•soundex(‘Ashcraft') == soundex(‘Ashcroft') == ‘A261’
•Soundex('Rupert') == soundex('Robert') == 'R163’
Element-level techniques: resource-based
•,%#A%50)A%&V'/:9&W%*
•𝐴 ⊑ 𝐵 '1*T*'5*)*"@:9$@8*9-*8#-9$@8*91*]
•apple ⊑fruit
•𝐴 ⊒ 𝐵 '1*T*'5*)*"@:#-$@8*9-*"929$@8 91*]
•animal ⊒cat
•𝐴 = 𝐵 '1*T*)$4*]*)-#*5@$9$@85
•student =pupil
•𝐴 ⊥ 𝐵 '1*T*)$4*]*)-#*)$39$@85*9-*3"#*5'I2'$%5*'$*3"#*:)-3*91*"'#-)-."@
•giant ⊥dwarf
Element-level techniques: resource-based
Sabou M., d’Aquin M., Motta E. (2008) Exploring the Semantic Web as Background Knowledge for Ontology Matching. In: Spaccapietra S. et al. (eds)
Journal on Data Semantics XI. Lecture Notes in Computer Science, vol 5383. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-92148-6_6
Background knowledge-based matching
http://www.semantic-web-journal.net/content/background-knowledge-ontology-matching-survey-0
T:A*'YA%&'Z).$?9:;#&'S#:<>%&?%'
M%A:;9.%AV
•/:9&W%*
•O9)#A>)*":#'6Q8A
•Z":-%&".)>':#*:>:?"%AG'%J?JG'
YT!,G'YZEMKWG'Z":Q:9*)>G'
T%,X
Structure-level techniques
•R9)B=50)A%&V'.:#A"&%9':#*:>:?"%A')A'>)0%>>%&'?9)B=AG')AA;-B*":#V'"C'
#:&%A')9%'A"-">)9'*=%#'*=%"9'#%"?=0:;9A'-;A*'0%'A"-">)9
•O)H:#:-@50)A%&V'.:#A"&%9':#>@'AB%.")>"A)*":#[?%#%9)>"I)*":#'9%>)*":#
•T:&%>50)A%&V'*)$%'"#*:')..:;#*'A%-)#*".'"#*%9B9%*)*":#':C'*=%'
:#*:>:?"%AG')AA;-B*":#V'"C'*<:'%#*"*"%A')9%'*=%'A)-%'*=%#'*=%@'A=)9%'
*=%'A)-%'"#*%9B9%*)*":#
•8#A*)#.%50)A%&V';A%'"#&"F"&;)>'9%B9%A%#*)*":#':C'.>)AA%AG'"J%JG'.>)AA%A'
)9%'.:#A"&%9%&'A"-">)9'"C'*=%@'A=)9%'-)#@'"#A*)#.%A
Structure-level techniques: graph-based
•P=">&9%#
•V<9*$9$a2#)1*5."#8)*#2#8#$35*)-#*53-&.3&-)22@*5'8'2)-*'1*3"#'-*'88#4')3#*
."'24-#$*5#35*)-#*"'%"2@*5'8'2)-
•!%)F%A
•V<9*$9$a2#)1*5."#8)*#2#8#$35*)-#*53-&.3&-)22@*5'8'2)-*'1*3"#'-*2#)1*5#35*)-#*
"'%"2@*5'8'2)-E*#=#$*'1*3"#'-*'88#4')3#*."'24-#$*)-#*$93
•8*%9)*"F%'C"H'B:"#*'.:-B;*)*":#
•i1*3"#*$#'%"I9&-5*91*3<9*$94#5*91*3"#*3<9*9$3929%'#5*)-#*5'8'2)-E*3"#@*<'22*I#*
89-#*5'8'2)-
Structure-level techniques: model-based
(%A.9"B*":#'!:?".A'1(!750)A%&
Structure-level techniques: instance-based
Wang S., Englebienne G., Schlobach S. (2008) Learning Concept Mappings from Instance Similarity. In: Sheth A. et al. (eds) The Semantic Web -ISWC 2008.
ISWC 2008. Lecture Notes in Computer Science, vol 5318. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-88564-1_22
Instance matching (record linkage)
Castano S., Ferrara A., Montanelli S., Varese G. (2011) Ontology and Instance Matching. https://doi.org/10.1007/978-3-642-20795-2_7
Value-oriented techniques
•6AA;-B*":#V'*=%'A"-">)9"*@'>%F%>':C'*<:'9%.:9&A'1%#*"*"%A7'.)#'0%'
&%9"F%&'0@'-)*.="#?'*=%'F)>;%A':C'*=%"9'.:-B)9)0>%')**9"0;*%A
•T:A*>@'C:.;A':#'A"-">)9"*@':C'A*9"#?')**9"0;*%A
•R")-).3#-aI)5#4Q*h4'3*H'53)$.#E*!8'3"a()3#-8)$*H'53)$.#E*j)-9 H'53)$.#
•Typographical variations, e.g. “organisation” vs “organization”
•V96#$aI)5#4Q*R95'$#*5'8'2)-'3@*9=#-*V;aiH;*=#.39-5*9-*<9-4*#8I#44'$%5
•Different conventions for describing data, e.g. “J.K. Rowling” vs “Rowling, J. K.”
•S"9$#3'.aI)5#4Q*!9&$4#KE*kl!ii!E*,#3):"9$#
•Phonetic similarity, e.g. “Kageonne” vs “Cajun”
Classification of matching vs non-matching
•/=%#'*=%'A"-">)9"*@':C'%).='B)"9':C'.:99%AB:#&"#?')**9"0;*%'F)>;%A'"A'
.:-B;*%&G')'&%."A":#'%#?"#%'"A'#%%&%&'*:'.>)AA"C@'<=%*=%9'*<:'
%#*"*"%A'-)*.=':9'#:*
•D#)-$'$%aI)5#4*
•!'8'2)-'3@aI)5#4
•0&2#aI)5#4
•R9$3#K3aI)5#4
Learning-based techniques
•,;B%9F"A%&'>%)9#"#?
•V-)'$'$%*4)3)Q*)*5#3*91*'$53)$.#*:)'-5*)$4*3"#'-*#K:#.3#4*.2)55'1'.)3'9$*A'B#B*
8)3."'$%*9-*$9$a8)3."'$%*-#.9-45G*
•k9$a3-'=')2*39*%#3*)*"'%"a7&)2'3@*)$4*I)2)$.#4*3-)'$'$%*4)3)*5#3
•,)$&)22@*)44'$%*)8I'%&9&5*.)5#5*'1*:955'I2#
•Y#A;B%9F"A%&'>%)9#"#?
•R2&53#-'$%*-#.9-4*:)'-5*<'3"*5'8'2)-*1#)3&-#5*3")3*I#29$%*39*3"#*5)8#*.2)55*A'B#B*
8)3."'$%*9-*$9$a8)3."'$%*-#.9-45G
•P:-0"#"#?'&"CC%9%#*'>%)9#"#?'*%.=#"D;%A
Similarity-based techniques
•P:#A"&%9"#?'*=%'"#B;*'9%.:9&A')A'>:#?')**9"0;*%'F)>;%A
•R9$.)3#$)3#*)22*)33-'I&3#*=)2&#5*'$39*9$#*5'$%2#*53-'$%
•6F%9)?%'A"-">)9"*@':C'%).='B)"9':C'.:99%AB:#&"#?')**9"0;*%'F)>;%A
•(#'%"3#4*)=#-)%#*8)@*I#33#-*-#12#.3*498)'$*6$9<2#4%#*
•O)$"#?'"#*:')..:;#*'*=%'C9%D;%#.@'%).='F)>;%':..;9A
•T*8)3."#4*mS)&2*!8'3"n*.9&$35*2#55*3")$*3"#*8)3."#4*m])8'4#2#*,#2'5'g<#n*
Rule-based techniques
•8#A*%)&':C'A"-">)9"*@'F)>;%AG')'Z::>%)#':;*B;*'"A')AA"?#%&'
•;9-*#K)8:2#E*'1*3<9*-#.9-45*4#$93'$%*I9965*5")-#*3"#*5)8#*=)2&#*9$*
)33-'I&3#5*mV'32#nE*mT&3"9-n*)$4*mS&I2'.)3'9$*@#)-nE*3"#-#*'5*)*=#-@*"'%"*
:-9I)I'2'3@*3")3*3"#*.9$5'4#-#4*-#.9-45*-#1#-*39*3"#*5)8#*I996B*
Context-based techniques
•O)$"#?'"#*:')..:;#*'*=%'9%>)*":#A="BA'0%*<%%#'9%.:9&A
•;9-*#K)8:2#E*8)3."'$%*.2&53#-5*91*-#.9-45*'$53#)4*91*'$4'='4&)2*-#.9-45
Optimisation
•M%&;.*":#':C'*=%'#;-0%9':C'.:-B)9"A:#A
•]29.6'$%Q*4'='4'$%*'$53)$.#5*'$*"989%#$#9&5*)$4*8&3&)22@*#K.2&5'=#*blocks
)$4*.98:)-'$%*9$2@*<'3"*'$53)$.#5*I#29$%'$%*39*3"#*5)8#*I29.6*
•!9-3#4*$#'%"I9&-"994Q*59-3'$%*'$53)$.#5*)..9-4'$%*39*3"#*=)2&#*3"#@*
)55&8#*9$*3"#*:-9:#-3@*<'3"*3"#*"'%"#53*4'5.-'8'$)3'$%*:9<#-*)$4*9$2@*
.98:)-'$%*'$53)$.#5*<'3"'$*)*5"'13'$%a <'$49<*91*)*1'K#4*4'8#$5'9$*
•M%&;.*":#':C'*=%'.:A*':C'%).='.:-B)9"A:#
•F$2@*)*5&I5#3*91*3"#*.9--#5:9$4'$%*)33-'I&3#*=)2&#5*8)33#-5
•E.g. title-author-year is a reasonable subset to distinguish books
Similarity filter and alignment extraction
•T)#@')>?:9"*=-A')9%'0)A%&':#'A"-">)9"*@':9'&"A*)#.%'.:-B;*)*":#J'6'
C%<':B%9)*":#A'.)#'0%'0)A%&':#'A"-">)9"*@[&"A*)#.%'-)*9".%AJ'
•\)9":;A'*=9%A=:>&"#?':B*":#A'*:'C">*%9'A"-">)9"*"%A
•f)-4*3"-#5"924*-#3)'$5*)22*3"#*.9--#5:9$4#$.#*)I9=#*3"-#5"924*$o
•H#23)*3"-#5"924*.9$5'535*91*&5'$%*)5*)*3"-#5"924*3"#*"'%"#53*5'8'2)-'3@*=)2&#*
9&3*91*<"'."*)*:)-3'.&2)-*.9$53)$3*=)2&#*4*'5*5&I3-).3#4o
•p):*3"-#5"924*-#3)'$5*3"#*.9--#5:9$4#$.#5*9-4#-#4*I@*4#.-#)5'$%*5'8'2)-'3@*
&$3'2*3"#*4'11#-#$.#*'$*5'8'2)-'3@*I#3<##$*3<9*.9--#5:9$4#$.#5*I#.98#5*
2)-%#-*3")$*$o*
•S-9:9-3'9$)2*3"-#5"924*.9$5'535*91*&5'$%*)5*)*3"-#5"924*3"#*:#-.#$3)%#*91*3"#*
"'%"#53*5'8'2)-'3@*=)2&#o
•S#-.#$3)%# -#3)'$5*3"#*$q*.9--#5:9$4#$.#5*)I9=#*3"#*93"#-5B*
Evaluation
Alignment quality measures
•];)>"*@'-%)A;9%A')9%'*=%'"#*9"#A".'-%)A;9%A':C'*=%')>"?#-%#*'1*=%@'&:'
#:*'&%B%#&':#')#@'9%C%9%#.%7V'
•3"-#5"924*9$*.9$1'4#$.#*9-*)=#-)%#*.9$1'4#$.#E*
•.9"#5'9$*8#)5&-#5*I#3<##$*8)3."#4*#$3'3'#5E*'B#BE*3"#'-*$#'%"I9&-5*)-#*
8)3."#4*<'3"*#)."*93"#-E*
•)8I'%&'3@*4#%-##E*'B#BE*:-9:9-3'9$*91*.2)55#5*8)3."#4*39*5#=#-)2*93"#-*.2)55#5E*
•)%-##8#$3*9-*$9$a4'5)%-##8#$3*I#3<##$*3"#*)2'%$#4*9$3929%'#5E*
•='92)3'9$*91*598#*.9$53-)'$35E*#B%BE*).@.2'.'3@*'$*3"#*.9--#5:9$4#$.#*:)3"5E*
•5)3'51).3'9$*91*5@$3).3'.*)$3'a:)33#-$5E*
•.9$5'53#$.@*)$4*.9"#-#$.#B*
Evaluation with a reference alignment
Euzenat Jérôme, & Shvaiko, P. (2013). Ontology matching (2nd ed.). Springer Berlin Heidelberg.
•C;/+1$1(. 4/3$,;/$2"!/2;3"1(2(82+(;;/+"'B28(,.)243"+!/$2T";,/2#($1"1I/$V2(I/;2"!/2"("3'2.,46/;2(82
;/",;./)243"+!/$2T";,/2#($1"1I/$23.)283'$/2#($1"1I/$V
•G/+3''24/3$,;/$2"!/2;3"1(2(82+(;;/+"'B28(,.)243"+!/$2T";,/2#($1"1I/$V2(I/;2"!/2"("3'2.,46/;2(82/j#/+"/)2
43"+!/$2T";,/2#($1"1I/$23.)283'$/2./73"1I/$V
•c*4/3$,;/2k2=2l2C2l2G2&2TC2m2GV
Evaluation without a reference alignment
•,)-B>"#? :9'B::>"#?50)A%&'%F)>;)*":#
•]&'24*)*5&I5#3*91*)2'%$8#$3*&5'$%*5)8:2'$%*9-*:992'$%
•hK:#-3*#=)2&)3'9$*
•E#&5*:5%#& %F)>;)*":#
•V#53'$%*3"#*7&)2'3@*91*)2'%$8#$3*&5'$%*)$*)::2'.)3'9$
•Q%9C:9-)#.% -%)A;9%A
•!:##4E*8#89-@E*5.)2)I'2'3@
Alignment lifecycle
The alignment life cycle
The matching methodology workflow
Learning and tuning
•T*8&2Ja53-)3#%@*2#)-$'$% )::-9)."*'5*&5#1&2*<"#$*5#=#-)2*2#)-$#-5*)-#*&5#4E*#)."*
9$#*")$42'$%*)*:)-J.&2)-*6'$4*91*:)r#-$*3")3*'3*2#)-$5*I#53B
•V&$'$% -#1#-5*39*3"#*:-9.#55*91*)4M&5J$%*)*8)3."#-*19-*)*I#r#-*1&$.J9$'$%*'$*
3#-85*91Q*
•I#r#-*7&)2'3@ 91*8)3."'$%*-#5&235E*8#)5&-#4E*#B%BE*3"-9&%"*:-#.'5'9$*9-*;a
8#)5&-#E*)$4*
•I#r#-*:#-19-8)$.# 91*)*8)3."#-E*8#)5&-#4*3"-9&%"*-#59&-.#*.9$5&8:J9$E*
#B%BE*#K#.&J9$*J8#E*8)'$*8#89-@B*
Summary
•X%*%9:?%#%"*@':C':#*:>:?"%A'"A'"#'*=%'#)*;9%':C'*=%'A%-)#*".'<%0^'
•K#*:>:?@'-)*.="#?'"A'B)9*':C'*=%'A:>;*":#^'
•8*'.)#'0%'0)A%&':#'-)#@'&"CC%9%#*'*%.=#"D;%A^'
•8*%9)*"F%'B9:.%AA'*:<)9&A'"-B9:F%&'B%9C:9-)#.%^
•O=%9%'9%-)"#'A%9":;A'.=)>>%#?%A')=%)&J'
Challenges
•!)9?%5A.)>%')#&'%CC"."%#*'-)*.="#?G
•T)*.="#?'<"*='0).$?9:;#&'$#:<>%&?%G
•YA%9'"#F:>F%-%#*G
•T;>*">"#?;)>"*@G
•,:.")>')#&'.:>>)0:9)*"F%'-)*.="#?G
•Y#.%9*)"#*@'"#'-)*.="#?G
•M%)A:#"#?'<"*=')>"?#-%#*AG
•6>"?#-%#*'-)#)?%-%#*J'
•JJJ'