Database Reference
In-Depth Information
9.4.2.2 Building Relations
After the identification of the entities set
ESN,
we use the existent relations in the
hypernode database (the relations that share
ESN
elements with other hypernodes or
among them) to find the set of relations
RSN
.
In order to facilitate this step, we have designed a set of patterns to apply this
kind of transformation to all the relations in the hypernode database. The pattern
will enumerate all the existing relations between persons by using only the hyper-
node database schema. After the relation pattern identification, we will search the
correspondent relations on the instances database.
A pattern relation
Pr
is defined by
Pr
such as
nPr
is the
name of the relation, hp
1
and hp
2
the hypernodes, which share the relation hyper-
nodes that represent people), and h
in
a mediator for this relation (the hypernode
used to identify the relation).
For each relation R
h
2
¼h
nPr
;
hp
1
;
hp
2
;
h
in
i
set of
LHD
relations, we check these conditions:
1. If R
h
:
ESN then h
s
or h
d
is added to
ESN
.
The relation “IS-A” allows us to find hidden entities that are not identified in the
previous step. In the relation construction process, we start by analyzing this
kind of relation to find in the next steps the relations related to the new
discovered entity.
2. If R
h
:
¼<
“IS-A”,h
s
,h
d
>
where h
s
or h
d
2
¼<
r,h
s
,h
d
>
where h
s
and h
d
2
ESN then two patterns are identified:
2.1 Pr
1
:
, if two entities (h
s
and h
d
) are already connected
in the
LHD
, we will search if their instances (h
is
and h
id
)are connected,
too.
2.2 Pr
2
:
¼<
r, h
is
,h
id
,null
>
j.
Pr
2
represents the relations between the instances of h
s
, which can be
connected with the same instance of h
d
.
3. If R
h
:
¼<
Same_hd.name, hp
i
,hp
j
,h
d
>
, where hp
i
¼
h
is
,hp
j
¼
h
is
, and i!
¼
¼<
rh
s
,h
d
>
, where h
s
2
ESN and h
d
2
ESN, then:
3.1 If r!
¼
“Part-of,” then pattern Pr
3
is extracted: Pr
3
:
¼<
Same_h
d
.name, hp
i
,
j. We search the h
s
instances
that are connected with the same instance of h
d
.
3.2 If r
hp
j
,h
d
>
, where hp
i
¼
h
is
,hp
j
¼
h
is
, and i!
¼
¼
“Part-of ” then the hypernodes that are “Part-of” h
d
are researched:
Firstly, for each h
j
2
}, a new
node is added to h
s
containing the name of h
j
and then the pattern Pr
4
is
extracted. Pr
4
:
{h\h has the relation R
h
:
¼<
“Part-of”, h
j
,h
d
>
h
s
.Pr
4
represents the relations between the instances of h
s
that share the same value
of h
j.
4. If R
h
:
¼<
Same_h
j
.name, hp
1
,hp
2
,h
j
>
, where hp
1
¼
h
s
,hp
2
¼
ESN, then:
4.1. a new node on h
s
containing the name of h
d
is added.
4.2. if h
s
has relations with other entities, then for each detected relations, a
pattern Pr
5
is extracted as Pr
5
:
¼<
rh
s
,h
d
>
, where h
s
2
ESN and h
d
2
¼<
Same_hs.name,hp
1
,hp
j
,h
d
>
, where
{e\e has relation with h
s
}.
By applying these patterns to our example, we can detect the relations between
the detected entities
Student
and
Director_thesis
:
hp
1
¼
h
d
and hp
j
2
Search WWH ::
Custom Search