Semantic search system using semantic ranking scheme
View Patent ↗A semantic search system using a semantic ranking scheme including: an ontology analyzer analyzing ontology data related to a search target to determine a weight value of each property according to a weighing method for property; a semantic path extractor extracting all the semantic paths between resources and query keywords and determining a weight value of each extracted semantic path according to the semantic path weight value determination scheme by using the weight value of each property; a relevant resource searcher traversing an instance graph of ontology based on a semantic path having a pre-set length and weight value of more than an expectation level to search resources that have a semantic relationship with the query keywords and are declared as a type presented in the query; and a semantic relevance ranker selecting a top-k results having the highest rank from among the candidate results extracted by the relevant resource researcher by using a relevance scoring function.
1. A semantic search system using a semantic ranking scheme comprising:
an ontology analyzer using a computer to analyze ontology data related to a search target to determine a weight value of each property wherein the ontology data comprises DBLP and IMDB data sets;
a semantic path extractor extracting all the semantic paths between resources and query keywords and determining a weight value of each extracted semantic path according to a semantic path weight value determination scheme by using the weight value of each property;
a relevant resource searcher traversing an instance graph of ontology based on a semantic path having a pre-set length and weight value of more than an expectation level to search resources that have a semantic relationship with the query keywords and are declared as a type presented in the query; and
a semantic relevance ranker selecting a top-k results having the highest rank from among the candidate results extracted by the relevant resource researcher by using a relevance scoring function;
wherein the relevance scoring function reflects the number of meaningful semantic paths between resources and keywords, the coverage of keywords and the distinguishability of keywords related to the resources;
wherein the relevance scoring function is represented by:
Rank
(
a
,
K
)
=
1
-
[
∑
1
≤
i
≤
K
(
D
(
k
i
)
·
(
1
-
NR
(
a
,
k
i
)
)
)
p
∑
1
≤
i
≤
K
D
(
k
i
)
p
]
1
p
wherein Rank(a,K) is a relevance scoring function with respect to a set of query keywords K of a resource ‘a’, wherein NR(a,k) is a normalized relation degree of resource ‘a’ with respect to a keyword k i which has a value greater than zero and less than or equal to 1, and wherein D(k i ) is subject discriminative power of the keyword k i and p is a tuning parameter for controlling AND-semantics;
wherein the NR(a,k i ) is calculated by:
NR
(
a
,
k
i
)
=
R
(
a
,
k
i
)
max
a
m
∈
A
R
(
a
m
,
k
i
)
for
k
i
∈
K
wherein R(a,k i ) is a relation degree of resource ‘a’ with respect to the keyword k i max R(α m ,k i ) is a maximum value of a relation degree, a m is a resource which makes R(α m ,k i ) a maximum value and ‘A’ is a set of resources;
and R(α,k i ) is calculated by equation shown below:
NR
(
a
,
k
i
)
=
R
(
a
,
k
i
)
max
a
m
∈
A
R
(
a
m
,
k
i
)
for
k
i
∈
K
wherein R(A,k i ) is a relation degree of resource ‘a’ with respect to the keyword k i max R(a m ,k i ), is a maximum value of a relation degree, a m is a resource which makes R(a m , k i ) a maximum value and ‘A’ a set of resources; and
R(a,k i ) is calculated by:
R
(
a
,
k
i
)
=
∑
ip
∈
IP
(
a
,
k
i
)
W
(
ip
)
,
for
k
i
∈
K
wherein IP(a,k i ) is a set of all the semantic path instances existing between the resource ‘a’ and the keyword k i , ‘ip’ is an element of the set ‘IP(a, ki)’ and W(ip) is a weight value of a semantic path instance ‘ip’;
wherein D(k i ) is calculated by:
D
(
k
i
)
=
irf
(
k
i
)
max
k
m
∈
K
irf
(
k
m
)
where k m is an element of the K and km makes irf(k m ) a maximum value
where irf(k i ) is calculated by:
irf
(
k
i
)
=
log
DV
DV
k
i
where DV comprises a set of data values on an ontology instance graph, and DV ki comprises a set of data values including a keyword k i .
2. The system of claim 1 , further comprising:
a database management system (DBMS) storing the weight value of each property determined by the ontology analyzer together with the ontology data.
3. The system of claim 1 , wherein a weighing method for property determines a property weight value according to discriminative power of a property and a predictability of a subject given the object, and vice versa, of a property.
4. The system of claim 3 , wherein the property weight value is determined by the equation shown below:
w ( p ( d,r ))=α· l ( p ( d,r ))+β· Ml ( p ( d,r ))
where w(p(d,r)) is a property weight value, I(p(d,r)) is a subject discriminative power of a property, MI(p(d,r)) is the predictability of the subject given the object, and vice versa, d is a domain, r is a range, and α and β are tuning parameters (0<α, β≦1).
5. The system of claim 1 , wherein the weight value of the semantic path is determined by the equation shown below:
W
(
sp
)
=
(
∏
p
(
d
,
r
)
∈
sp
w
(
p
(
d
,
r
)
)
)
·
δ
length
(
sp
)
-
1
where W(sp) is a weight value with respect to a semantic path, sp is a semantic path, w(p(d,r)) is a property weight value, length(sp) is the number of properties constituting a semantic path, and δ is a tuning parameter reflecting a penalty over the length of a semantic path (0<δ≦1).