INFORMLEDGE SYSTEM

A Modified Knowledge Network with Autonomous Nodes using Multi-lateral Links

T.R. Gopalakrishnan Nair

Dayananda Sagar Institution, 4

Floor, New Business Block, KumaraswamyLayout, Bangalore, India

Meenakshi Malhotra

D. S. Institution, 4

Floor, New Business Block, KumaraswamyLayout, Bangalore, India

Keywords: Informledge System (ILS), Knowledge Network Node (KNN), Multi-lateral links, Link Manager.

Abstract: Research in the field of Artificial Intelligence is continually progressing to simulate the human knowledge

into automated intelligent knowledge base, which can encode and retrieve knowledge efficiently along with

the capability of being is consistent and scalable at all times. However, there is no system at hand that can

match the diversified abilities of human knowledge base. In this position paper, we put forward a theoretical

model of a different system that intends to integrate pieces of knowledge, Informledge System (ILS). ILS

would encode the knowledge, by virtue of knowledge units linked across diversified domains. The proposed

ILS comprises of autonomous knowledge units termed as Knowledge Network Node (KNN), which would

help in efficient cross-linking of knowledge units to encode fresh knowledge. These links are reasoned and

inferred by the Parser and Link Manager, which are part of KNN.

1 INTRODUCTION

Enormous amount of demand exists in Artificial

Intelligence and cognitive systems to facilitate

knowledge storage and retrieval with consistency

and scalability. As from Weishan Zhang, Thomas

Kunz (2006), certain amount of research has

already taken place that helps in systematically

connecting words to formulate sentences of

admissible meanings. Though the efforts are

continued in this field, no brakethrough has been

reported for evolution of an auto routed knowledge

component that is not only capable of storage, but

can intelligently parse and link. Informledge System

transforms information stored into meaningful

knowledge by virtue of multi-lateral links which are

inferred and reasoned by parser and link manager.

This paper is organized as follows. Section 2

discusses Research Background. Section 3

describes Research Objective and Section 4 gives

composition of Knowledge Network Node. Section

5 gives the Conceptual Model of ILS. Section 6

shows how knowledge is encoded into ILS and

other knowledge representations. Finally, we

conclude in section 7.

2 RESEARCH BACKGROUND

Since early 1980s ontologies have been vital part

for knowledge representation in various fields of AI

and Semantic Web. The term ontology which

originated from the field of philosophy is meant to

represent what exist. In computer science theory,

“ontology is formal, explicit specification of a

shared conceptualization” (Thomas R. Gruber,

1993). For knowledge sharing and reuse, numerous

representations have been devised to structure the

stored information.

2.1 Knowledge Bases

The need for the systems to encode and retrieve

information intelligently has lead to the creation of

knowledge base systems. Knowledge comprises of

concepts, theory, facts and rules which are modelled

using ontologies. The necessity to integrate domain

specific ontologies and reuse data elements from

existing system has led towards Standard Upper

Ontology (SUO). SUO provide definitions for

general concepts at higher-level not including

domain specific concepts but acts as a foundation

351

Gopalakrishnan Nair T. and Malhotra M..

INFORMLEDGE SYSTEM - A Modiﬁed Knowledge Network with Autonomous Nodes using Multi-lateral Links.

DOI: 10.5220/0003069103510354

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 351-354

ISBN: 978-989-8425-29-4

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

for more specific domain ontologies (Ian Niles,

Adam Pease, 2001). The upper ontologies available

are: OpenCyc, BFO, DOLCE and DnS , GFO

,IDEAS ,SUMO , IDM , Biomedical ontology ,

COSMO.

However, Ontology development faces new

challenges as listed below (Ontology Development

Pitfalls, 2005, para 1):

 Development of domain ontologies has led to

semantic heterogeneity (Robert M. Colomb,

2007, ‘Upper Ontologies’, para 2).

 It fails to distinguish between different

relationships like ‘instance-of’ relationship.

 Problem in reasoning the events that are events

relations between concepts

 Failure to model change in facts over time.

2.2 Semantic Web

The ever increasing content on World Wide Web

requires it to be shared and reused, which gave way

to the development of Semantic Web (

Ivan Herman,

2010). According to Sean Bechhofer, Ian Horrocks

and Peter F. Patel-Schneider (2003) Semantic Web

require:

 Metadata to access the shared information,

 Ontologies to provide vocabulary for

annotations

 Standard web ontology language to have a

common syntax before we can share semantics

Semantic Web imposes new challenges as listed

by Kenneth, Paulo C. G. Costa, Mieczyslaw M.

Kokar, Trevor Martin, Thomas Lukasiewicz (2008):

 Pages must be semantically annotated

through processes that are mostly manual and

require good engineering skill.

 Generating metadata is the requirement.

 People's privacy could become compromised.

 Logical contradictions, inconsistencies,

which will inevitably arise when ontologies

from separate sources are combined.

 It extracts information mainly from web

which is typically incomplete and uncertain.

 There is a risk of disingenuous information,

as anyone can publish anything on the web.

3 RESEARCH OBJECTIVE

The proposed system (ILS) aims at formulating

knowledge through the intelligent links. The links

form the vital part of this system, which is not just

an arc but is a detailed specifier for the linkage i.e.

having information about type, direction,

connectivity, source and destination nodes, creation

time, usage, etc. of the link. It can be a connection

between two nodes/ concepts/ entities/ events/

elements or even two links.

The system is knowledgeable by virtue of its

understanding i.e. it can classify, correlate and

extrapolate the information stored using the

linkages. ILS provides knowledge-to-knowledge

interaction which is not possible if we are storing

merely text. This interaction is possible with multi-

lateral linking. Ontologies represent concepts, but

concepts change with time based on occurrences of

some events. In the proposed system, any event or a

change in a fact is a knowledge evolution and is

handled via link creation, modification or deletion.

4 KNOWLEDGE NETWORK

NODE

ILS encompasses set of autonomous nodes termed

as Knowledge Network Node (KNN). KNN will

help in automatically encoding new knowledge

within the existing ILS with the help of its four

quadrants (shown in Figure.1):

Figure 1: Knowledge Network Node-

 Input / Output: This part has the knowledge

to be encoded and the identified knowledge

units as input and the next KNN as output.

 Storage and Semantic Parser: It stores the

knowledge unit and its meaning. In addition

to this it has a parser which is responsible for

generation of link which is reasoned and

inferenced by the parser itself.

 Link Manager: It helps in addition, deletion,

classification and prioritization of the links.

Prioritization of links is based on many

factors such as: links connecting KNN in

demand, recently used links, etc.

 Link Database: It holds all the information

about the links of encoded knowledge. It

stores link attributes i.e. link description,

type: additive/integrative/inclusive, creation

time, source, destination, last used, status, etc.

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

352

During knowledge retrieval, the existing links

helps in formulation of knowledge, from the

destination node, an instance of this thread can be

easily pulled out as shown in Figure 2. This link

information will also help in faster encoding of new

knowledge, by using existing KNNs.

Figure 2: Extraction of knowledge thread.

5 INFORMLEDGE SYSTEM

Informledge System (ILS) is a collection of linked

KNNs. ILS targets at encoding knowledge

meaningfully and with less abrasion, to make the

retrieval meaningful and easy. ILS has Knowledge

Library which consists of an index of all KNNs

present with the system. Knowledge Library serves

as an entry into ILS. Encoding can start from any

arbitrary KNN termed as KNN1 and this is

identified by Knowledge Encoder (KE) which is a

part of ILS architecture, as shown below in Figure

3. The first knowledge unit can be considered as the

head of the knowledge from where we splinter to

the subsequent nodes through intelligent links.

Figure 3: Conceptual Model of Informledge System.

The knowledge unit extracted by the KE is

fetched and knowledge to be encoded is provided as

input to KNN1. Semantic Parser at KNN1 parses

the knowledge and gets the links to be established

between this and next KNN. Link Manager appends

links to the database, thus connecting the 2 KNNs.

The Parser at KNN1 creates output by appending

the address of next KNN with the input, which is

then taken as input to the subsequent KNN. Next

KNN is then fetched from the knowledge library

and the same procedure is repeated for the next

KNNs also. This process goes on till the knowledge

is encoded completely i.e. all the knowledge units

are linked. Thus encoding process advances from

one to other KNN, through the ingrained

intelligence of links. Links in ILS carries

information to the destination KNN that parse this

link information leading to creation of links to

subsequent nodes. Knowledge of ILS will evolve

with increase in number of links, resulting from

more and more knowledge being encoded into ILS.

Also, new knowledge can be encoded using existing

KNNs and freshly created links.

6 ENCODING

Efficient knowledge retrieval is based on how we

encode knowledge and how we represent

knowledge.

6.1 Encoding into Existing KBs

Ontologies are used to represent concepts. Classes

are used to represent concepts and collections,

where an instance of a class represents individuals,

and attributes represent individual or class

properties. Along with classes, class

interconnections, assertions, rules and restrictions

are part of any ontology (Héctor Díez-Rodríguez

and Guillermo Morales-Luna, José Oscar Olmedo-

Aguirre, 2008, ‘3.2’ Sec, para 2). To encode

something like “African Lions are Strong” Cyc

Ontology will need to create terms or classes for

Africa and Lion. OWL builds on RDF and RDF

Schema which adds more vocabulary for describing

properties and classes, relations between classes

(e.g. disjointness), cardinality (e.g. "exactly one"),

equality, richer typing of properties, characteristics

of properties (e.g. symmetry), and enumerated

classes (Ivan Herman,2007 W3C).

6.2 Encoding into Semantic Web

According to Sean Bechhofer, Ian Horrocks and

Peter F. Patel-Schneide r (2003) Semantic Web uses

URI on web as in Figure 4, to link datasets like

Dbpedia, GeoNames, FOAF, VIAF, Freebase etc.

Figure 4: Data linkages in Semantic web.

INFORMLEDGE SYSTEM - A Modified Knowledge Network with Autonomous Nodes using Multi-lateral Links

353

To encode Africa Lions are Strong, the

properties of entities need to be updated along with

creation of new data and thus linking it. For this we

need to create semantic web page for lion and

define a class for it. Then need to define properties

like to which category they belong and where they

belong and linking the same to the page of the sub-

category of lion, carnivore of Africa. The

knowledge about African Lion being strong is part

of the value of abstract property of this page.

6.3 Encoding into ILS

To begin with, the knowledge “African Lions are

Strong” is given as input to KE, which screens it

and recognizes the sub-knowledge units as: Africa,

Lion and Strong, represented as KNN1, KNN 2 and

KNN3 as shown in Figure 5. The three KNNs

belong to three different domains, but the structure

used to represent the three is the same. From KNNs

the links are there to n-number of other knowledge

nodes, forming a cloud of knowledge.

Figure 5: Encoding into ILS depicting knowledge Clouds.

The link between KNN1 and KNN2 depicts that

the two sub knowledge entities are connected,

which means knowledge about lions in Africa and

Africa having lions both are there. Moving further,

if KNN2 and KNN3 are not linked i.e. the

knowledge African lion being strong is not there.

Then the Link manager at KNN2 would create a

link and would update the link database at the

respective KNN’s i.e. KNN1, KNN2 and KNN3.

These links implies that Africa, Lion and Strong are

now linked. Thus when both the links are taken

together, the following knowledge could be

retrieved: “African Lions are strong”, “Lions of

Africa are strong” and “Strong Lions are in Africa”.

Semantic web will not be able to link back to

same thing i.e. Strong to African Lion, unless

another entity for strong is made. And in other

knowledge bases the reverse assertion need to be

inserted. However, using the link properties in ILS,

we extract the knowledge thread starting from KNN

‘strong’ and linking it to ‘Lion’ belonging to

‘Africa’. So far knowledge is fed manually into ILS

but to simulate real human knowledge, we could

use Natural language processing and domain

experts. It shows a scalable model with high degree

of rationality.

7 CONCLUSIONS

Informledge System deals with linked knowledge as

a whole, rather than just connected words, which

could be later extracted for a purpose. ILS combines

the essential units of a KB i.e. words and logic, into

KNN and its multi-lateral links reaching wider

scopes that are not available today. ILS works fairly

well with limited number of KNNs however it is

required to simulate the system with real world

model and need to couple it with mammoth amount

of KNNs linked across domains to handle the

knowledge explosion. In addition to this, advanced

studies of Tensor in vector space of multiple nodes

and links is under investigation to achieve further

progress in managing multi-lateral links. The future

work includes analysis of link properties along with

its comparison to the biological properties of

neurons, which would provide more insight to the

knowledge handling capability of the brain.

REFERENCES

Adam Pease, 2010. Ontology Development Pitfalls.

Retrieved October 3, 2005, from http://www.

ontologyportal.org /Pitfalls.html

Héctor Díez-Rodríguez and Guillermo Morales-Luna,

José Oscar Olmedo-Aguirre, 2008. Ontology-based

Knowledge Retrieval. Seventh Mexican International

Conference on Artificial Intelligence

Hendler, James, Berners-Lee, Tim and Miller, Eric,

October, 2002. Integrating Applications on the

Semantic Web. Journal of the Institute of Electrical

Engineers of Japan.

Ian Niles, Adam Pease, 2001. Towards a Standard Upper

Ontology. 2nd International Conference FOIS-2001.

Ivan Herman, 2007. Web Ontology Language, W3C.

Retrieved 15 October, 2007 from http://www.

w3.org/2004/OWL/#ontologies

Ivan Herman, 2010. W3C Semantic Web Activity.

Retrieved May, 2010 from http://www.w3.org /2001 /

sw /

Robert M. Colomb (2007) Ontology And The Semantic

Web ISBN 978-1-58603-729-1 IOS Press, 2007.

Thomas R. Gruber, 1993. A Translation Approach to

Portable Ontology Specifications. In: Knowledge

Acquisition. 5: 199-220.

Weishan Zhang, Thomas Kunz, 2006. Product Line

Based Ontology Development for Semantic Web

Service. Proceedings of the Second IEEE

International SOSE'06.

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

354