A Meta Description Logics Knowledge Base

Published Date: 02 Nov 2017

FOR ARABIC LANGUAGE PROCESSING

Yasser YAHIAOUI1, Ahmed LEHIRECH2

1. Centre Universitaire de N�ama, 2.Universit� Djillali Lyebes SBA.

1. [email protected], 2. [email protected]

ABSTRACT

Description Logics (DLs) are

successful knowledge representation

formalisms, which can be used to represent the

terminological knowledge of an application

domain in a structured and formally wellunderstood

way. They are employed in various

application domains, such as natural language

processing, configuration, and databases .Sf.

In this work we try to use this

performance of DL systems to describe a Meta

knowledge base defining the classes of the

word in Arabic language and relation between

them. This can be useful for syntactic

categorisation of sentences which is very

important for automatic language processing.

KEYWORDS

Description logics, Arabic language

processing, knowledge representation, Meta

knowledge base.

INTRODUCTION

From the linguistic point of view, the

language processing involves distinct levels

namely: the lexical level, syntactic level, the

semantic and pragmatic level. All these levels

contribute to the representation of semantic

information. Thus, the role of lexicology

becomes very productive since the meaningful

representation of the lexicon is provided, and

this helps the syntactic and semantic rules to

become easier. Dixon 1991 in his work on

syntax-based semantics, has proved that the

irregularities and idiosyncrasies can be

predicted from the semantics of words [01].

This aspect of idiosyncrasy (meaning

non-regular behaviour) leads us to think of a

heuristic treatment in order to manipulate these

syntactic elements which can drive to a kind of

classification. The latter facilitates the isolation

and segmentation of the sentences in a text

according to their meaning as a criterion of this

classification.

The powerful knowledge

representation formalism of description logics

is providing an expressive tool which can be

useful to create a Meta-knowledge base. This

part of the knowledge base can be used as an

add-on for ALP systems. It can also be

considered as a tool for searchers on the Arabic

language processing that provides a description

for the totality of the Arabic words.

The description can be summarized in

creating hierarchy classes of the Arabic word

that provides a dependency graph based on the

ISBN: 978-0-9853483-3-5 �2013 SDIWC 254

subsumption relationship, and in this way we

define the first part of the KB which we called,

at this level, the Meta-Box and by the totality

of words existing in the Arabic Word-Net

ontology we define the second part including a

terminological and assertional Box.

MOTIVATION

In the Beginning, it is important to

understand the notion of the natural language

processing (NLP) which is defined in literature

as: �[....] a theoretically motivated range of

computational technique for analyzing and

representing naturally occurring texts at one or

more levels of linguistic analysis for the

purpose of achieving human-like language

processing for a range of tasks or

applications�[02]. This definition leads us to

think about the levels of representation and

especially on the syntactic categorisation

which can carry the roles and definitions to

construct meaningful representation of each

word within the language.

For the Arabic language, the word has

a role in a sentence and this role can be

detected by the classification of the word

syntactically, the fact which guides us to think

about well-understood way of representations.

Since the description logics are regarded as a

structured and formally well-understood way

that uses representation languages as the KLOne;

�Since the days of the KL-One system,

one of the main application of description logic

has been for the semantic interpretation in

natural language processing� [03].

We aim in this representation at

facilitating the semantic interpretation inspired,

inthis way, by the fact that �semantic

interpretation is the derivation process from the

syntactic analysis of utterance to its logical

form�[03]. The lexical part becomes more and

more important because we can begin the

semantic integration from the lexical level by

introducing a lexical semantic which means the

specification of the semantic of each concept.

The use of DLs formalism to describe

a Meta knowledge base using the syntactic

roles of the Arabic word aims at constituting

the lexical semantic part of the knowledge base

which is not newly utulized; since we find, for

instance, in the literature �a part of the

knowledge base constitutes the lexical

semantic knowledge, relating words and their

syntactic propertied to concept structures [...],

giving a deep meaning to concept� [03].

THE ARABIC LANGUAGE

PROCESSING (ALP) SYSTEM

ARCHITECTURE

The architecture we propose here is

basically constituted of two parts, the first one

contains the high level knowledge giving the

lexical and syntactical semantics. The second

one gives the ordinary description of concepts

(words) constructed as a KB built from the

Word-Net Arabic ontology.

ISBN: 978-0-9853483-3-5 �2013 SDIWC 255

Fig 01: System Architecture

In the traditional DLs systems �A

knowledge base (KB) comprises two

components, the TBox and the ABox. The

TBox introduces the terminology, i.e., the

vocabulary of an application domain, while the

ABox contains assertions about named

individuals in terms of this vocabulary� [04].

However; the addition of the Meta-

Box in the present system (fig 01) can be

considered as a consolidation for the traditional

system. This fact makes the introduction of the

syntactic role of the term (concept) with its

definition possible and ti is implicitly defined

in the class of the word which is integrated by

establishment of the link between the Meta-

Box and the T-Box using the relation of the

subsumption. Since all the terms existing in

Word-Net Arabic ontology can be a subsumed

by a concept among the Meta-Box concepts

either directly or indirectly.

The T-Box / A-Box are traditionally

constructed by the translation of definitions in

the Ontology of concepts and their relations to

DLs knowledge base defined using the KLOne

description language.

THE WORD IN THE ARABIC

LANGUAGE

The word in Arabic language can be

viewed as an occurrence of a node of a

dependency graph representing a hierarchical

organization of the classes existing for the

syntactic roles as following : the word is called

in Arabic �Kalima�, this model includes the

three nodes which are �Fiil� (verb) , �Ism�

(noun) , and �harf� (propositions, conjunctions

and so on). And all of these terms have

dependency with others to create the hierarchy

of concepts. (see fig 02)

Word-

Net

Arabic

Translation

T-Box

A-box

Meta-Box

Subsumption

links

Global T-BOX

ISBN: 978-0-9853483-3-5 �2013 SDIWC 256

Fig 2: Hierarchy of Syntactic Roles of Words in Arabic Language

This hierarchy have been inspired from

a lecture in the Arabic documentations

especially the poems of el ADJROUMIA and

ELFIAT IBN MALEK; but for the sake of

semantic implementation we have focused on

the classes which influence the meaning and

the use of the word in sentences. In order to

realize this graph in fig 02 we use the same

techniques of the dependency graph

construction known as a graphic representation

which is �a sample dependency graph in which

word nodes are given in bold face and

dependency relations are indicated by labelled

edges�[05].

In this proposition the relation

represented by edges is �IS A� which is

interpreted as subsumption in DLs terminology

or subset in the interpretation of DLs in set

theory seeing the concepts as a set of

individuals.

THE REPRESENTATION AS DLS

KNOWLEDGE BASE

The graph in the fig 02 here can reflect

the concepts of the Meta-Box and the relation

between them, the thing that helps in defining

the axioms and constructing the Meta-Box

which is the essential part because of the

meaningful representation given implicitly

with the specification of the syntactic roles of

each word defined in the second part. We

present here the description in the KL-One

language written and constructed for the

definition of this first level part of the global

knowledge base

ISBN: 978-0-9853483-3-5 �2013 SDIWC 257

fig 03: A Sample of the Meta-Box Described in KL-One

CONSTRUCTION OF THE SECOND

LEVEL

The relations existing in the linguistic

Word-Net Arabic are known and limited in an

ultimate number. We find eleven (11) relations

namely: Antonym, synonym, Meronym

(inverse of holonym), hyponym, implication

(entailment), causality, value, has as value,

affinity, derived from, similar to. All of these

relations can be interpreted By DLs language

using the known constructors and quantifiers

and we propose these interpretations as

following: [6, 7, 8]:

Antonym: in DLs it becomes disjunction

Synonym: in DLs we replace it by equivalent

Meronym: interpreted as a restriction or role.

Hyponym: X is a hyponym Y interpreted X is

subset of Y or X is subsumed by Y.

Implication: Y imply X means logically

disjunction between negation(Y) and X

(Neg(Y) U X)

Causality: using roles

Value: assertion

Has As Value: assertions

Affinity: restrictions

Derived From: Y derived from X means

logically disjunction between Negation (X)

and Y (Neg(x) U Y).

Similar To: equivalence

We note here, in this second level, that

we have to define two parts: the first is

terminological and the second is assertional but

we can say that there is causality between the

interpretations of the relation existing and the

position of each possible word defined in

Word-Net Arabic in each one of the two parts.

And this is clear because the word is well

ISBN: 978-0-9853483-3-5 �2013 SDIWC 258

defined if we can represent all its relations

linking the current word with all the others, in

another way, in DLs terminology the concept

is well defined if we reach a level of

representation in which all relationship with

others concepts are represented that means the

encapsulation of all the elements of the set of

individuals concerned by the definition of the

concept.

LINKS BETWEEN THE TWO

LEVELS

The classes defined in the Meta-Box

are the collection of what can an Arabic word

means because it is a high level categorization

of the terms and in this way, we can be sure

that all concepts existing in the TBox of the

second level are included in one at least of the

bottom concepts of the Meta-Box and it can be

considered as an assertion in this level; but

globally, the Global Terminological Box GTBox

is divided into two part the Meta-Box and

the T-Box. However; the Global Assertional

Box is similar to the assertional Box (see fig

01). We mathematically represent all this as:

CONCLUSION AND PERSPECTIVES

In this work, we arrived to evaluate an

idea which is new in the domain of Arabic

language processing and especially for the DLs

based systems. This idea consist to create an

add-on which we named the Meta-Box for the

purpose of getting richer lexicology integrating

syntactic roles with the definition of every

word existing in the Arabic linguistic ontology

named Word-Net.

As results we define a new architecture

for ALP description logics based system with

enlarged knowledge base in where we have the

traditional representation consolidate with

other knowledge integrated to enhance

efficiency, expressivity and the use cases of

this Global KB. This letter is designed to be a

useful tool for all kinds of treatments and

manipulations existing in the ALP domain as

the question answer systems, the automatic

translation, the text summarization, SF.

We propose to evaluate the GKB an

automated sentences segmentation which can

provide a lot of efficiency for the ALP

applications especially for the case of

automated translation and text segmentation.