Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová,...

21
Semantic Network as Semantic Network as Continuous System Continuous System Technical University of Košice Technical University of Košice d d oc. Ing. Kristína Machová, oc. Ing. Kristína Machová, PhD PhD . . Ing. Stanislav Dvorščák Ing. Stanislav Dvorščák WIKT WIKT 2010 2010

Transcript of Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová,...

Page 1: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

Semantic Network as Semantic Network as Continuous SystemContinuous System

Technical University of KošiceTechnical University of Košice

ddoc. Ing. Kristína Machová, oc. Ing. Kristína Machová, PhDPhD..

Ing. Stanislav DvorščákIng. Stanislav Dvorščák

WIKTWIKT 2010 2010

Page 2: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

22

MotivationMotivation Semantic network Semantic network Discrete versus continuous network Discrete versus continuous network ImplementationImplementation ConclusionsConclusions

ContentContent ContentContent

Page 3: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

33

Quantity of web information cannot be processed by people.

Key search: based on key words or phrases results are depended on used vocabulary search in prepared index recalls web pages instead of searched information recall and precision are usually low

MotivationMotivation MotivationMotivation

Page 4: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

44

MotivationMotivationMotivationMotivation

Solution?Solution? Web search which is provided automatically by Web search which is provided automatically by

machines (software). machines (software). Machines can process bigger amount of information.Machines can process bigger amount of information. Semantic search:Semantic search:

is based on semantic understanding of searched is based on semantic understanding of searched information and contextual informationinformation and contextual information

uses semantic technologies (metadata, ontologies, uses semantic technologies (metadata, ontologies, nonmonotonic logics, semantic agents, semantic nonmonotonic logics, semantic agents, semantic networks…)networks…)

Page 5: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

55

Semantic networkSemantic networkSemantic networkSemantic network

Why?Why? Semantic technologies (RDF, OWL…) represent Semantic technologies (RDF, OWL…) represent

higher standards. higher standards. But they need the creation of parallel web to current But they need the creation of parallel web to current

heterogeneous web (HTML documents).heterogeneous web (HTML documents). Semantic network can represent documents in various Semantic network can represent documents in various

forms (HTML, RDF, OWL,…):forms (HTML, RDF, OWL,…): our implementation uses Time Delay Neural our implementation uses Time Delay Neural

Network (TDNN)Network (TDNN) is enriched by contextual informationis enriched by contextual information

Page 6: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

66

Semantic network Semantic network Semantic network Semantic network

In our implementation: digraf which contains nodes and edges represents

SN nodes represent symbols (pieces of information);

have their own validity/invalidity edges are used to signal transitions; are oriented

from start points to end points signal is a real value from the range <-1,1>

Page 7: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

77

Semantic Network Semantic Network Semantic Network Semantic Network

The signal propagation in numerical domain:

Time delay of given signal (-1, -2, -3 ) causes

that symbol “2” became that symbol “2” became valid before symbol “x” valid before symbol “x” and it became valid and it became valid before symbol “4”.before symbol “4”.

The mathematic equation 2 x 4 + 1 = 9 can be represented by the rule: p+(px(2, 4), 1) = 9

Page 8: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

88

Semantic Network Semantic Network Semantic Network Semantic Network

The signal propagation in real estate domain:

Time delay of given Time delay of given signals (-1, -2) causessignals (-1, -2) causes

that “garage” (more that “garage” (more important demand for important demand for user) is fired before user) is fired before “1.floor”. “1.floor”.

The signal propagation in text document domain:The signal propagation in text document domain:Set of letters can be declared as a word only if these Set of letters can be declared as a word only if these letters became valid in the given order (sequence).letters became valid in the given order (sequence).For example: “before” “eforeb”For example: “before” “eforeb”

Page 9: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

99

Semantic network Semantic network Semantic network Semantic network

The signal propagation in text document domain:

Page 10: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1010

Semantic network Semantic network Semantic network Semantic network

The signal propagation in text document domain:

Approximately 100 squire metres?What does it means?100-x or 100+x ?

50x2,5 – do not understand 125 – understandmachine deduces

Page 11: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1111

Discrete versus continuous Discrete versus continuous networknetwork Discrete versus continuous Discrete versus continuous networknetwork

Continuous semantic network represents one block of the “Intelligent Search Machine with the Semantic support” :

IM framework – variables of the program are continuous information (not discrete data)

semantic network “lives” some time and hidden knowledge is deduced

semantic network is a dynamic network (recurrent, heterogeneous)

Some of the nodes are program modules (e.g. mathematic module).

Page 12: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1212

Discrete versus continuous Discrete versus continuous networknetwork Discrete versus continuous Discrete versus continuous networknetwork

Variables as discrete information:

At the time of calculation, the discrete values of variables are considered and the result is a discrete value.

Information z=x+y is forgotten after the discrete time of calculation.

int x = 2;int y = 3;

// z is 5int z = xPlusY(x, y);x= 4;

// z is still 5…private int xPlusY(int x, int y) { return x+y;}

Page 13: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1313

Discrete versus continuous Discrete versus continuous network network Discrete versus continuous Discrete versus continuous network network

IntDelta x = new IntDelta(2);IntDelta y = new IntDelta(3);IntDelta z = xPlusY(x, y); // z is 5z.getValue() returns actual valuex.setValue(4);// z is 7z.getValue()…private IntDelta xPlusY(IntDelta x, IntDelta y) { return IntDelta.plus(x, y);}

Variables as continuous information: Z continually changes its value during the change of any of the arguments of the sum. Z does not represent static value but information z=x+y.

Page 14: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1414

Sampling of the new valueSampling of the new value Sampling of the new valueSampling of the new value

The implementation cannot avoid the sampling The implementation cannot avoid the sampling because:because:

1)1) semantic network works continuouslysemantic network works continuously2)2) computer is a discrete machine.computer is a discrete machine.An operation can be dependent on time or not:An operation can be dependent on time or not: operation independent on time operation independent on time – considered value – considered value

can be actualized at the time of accesscan be actualized at the time of access operation dependent on time operation dependent on time – considered value – considered value

must be actualized continuouslymust be actualized continuously pragmatic accesspragmatic access – using of the time window (5 last – using of the time window (5 last

values, comparable precision).values, comparable precision).

Page 15: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1515

ImplementationImplementation ImplementationImplementation

The current system version has these four main parts:The current system version has these four main parts:1)1) harvester (document collector)harvester (document collector)2)2) document storedocument store3)3) semantic network generatorsemantic network generator4)4) inference mechanism (browser and reasoner) inference mechanism (browser and reasoner) It is realized via 3-layer standard J2EE architecture:It is realized via 3-layer standard J2EE architecture: DAO LayerDAO Layer Business Layer Business Layer Presentation Layer Presentation Layer ..

Page 16: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1616

ImplementationImplementation

DAO (Data Access Object) Layer

is ORM framework using IBATIS and MySQL database.

ORM – Object Relation Mapping

IBATIS – bridge betweenobjects and databases

RDBMS – Relational Data Base Management System

Page 17: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1717

Implementation Implementation

Business Layer is realized via framework IoC (Inversion of Control

container). It realizes the implementation of SOA (Service

Oriented Architecture). It has logging and transaction realized by AOP –

Aspect Oriented Programming.

Presentation Layer presents generalization of a markup which is realized

by component oriented framework.

Page 18: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1818

ConclusionsConclusions ConclusionsConclusions

The possibility of the implementation of dynamic The possibility of the implementation of dynamic systems.systems.

Novelty of the idea is: Novelty of the idea is: using TDNN (computational intelligence) to using TDNN (computational intelligence) to

meet the vision of the semantic webmeet the vision of the semantic web using fuzzy - cognitive map for searching and using fuzzy - cognitive map for searching and

reasoning.reasoning. For the future: For the future:

extension of the implementation with the transactions to extension of the implementation with the transactions to avoid the mistakes during the connections of net’s nodes avoid the mistakes during the connections of net’s nodes to other resources.to other resources.

Page 19: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

1919

ConclusionsConclusions ConclusionsConclusions

The implementation:The implementation: has valid architecture which enables easy has valid architecture which enables easy

extension and next developmentextension and next development works correctly in a numeric domainworks correctly in a numeric domain achieves interesting results in the domain of text achieves interesting results in the domain of text

documents (similarity of words, resistance against documents (similarity of words, resistance against typos)typos)

For the future: For the future: design and realization of the extension based on design and realization of the extension based on

ontology.ontology.

Page 20: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

2020

Thank you for your attentionThank you for your attention

Page 21: Semantic Network as Continuous System Technical University of Košice doc. Ing. Kristína Machová, PhD. Ing. Stanislav Dvorščák WIKT 2010.

2121

ImplementationImplementation

[email protected]

- implementation details