Arabase
A Database Combining Different Arabic Resources with Lexical and Semantic
Information
Hazem Raafat
1
, Mohamed Zahran
2
and Mohsen Rashwan
3
1
Computer Science Department, Kuwait University, Kuwait City, Kuwait
2
Computer Engineering Department, Cairo University, Giza, Egypt
3
RDI, Faculty of Engineering, Cairo University, Giza, Egypt
Keywords: Natural Language Processing, Arabic Language Resources, Text Similarity Matching, Arabic Language
Resources Integration.
Abstract: Language resources are important factor in any NLP application. However, the language resource support
for Arabic is poor because the existing Arabic language resources are either scattered, inconsistent or even
incomplete. In this paper we discuss the notion of having an integrated Arabic resource leveraging various
pre-existing ones. We present a comparison between these resources then we present preliminary fully and
semi-automated methods to integrate these resources. This work serves as a bootstrapping for a rich Arabic-
Arabic resource with a good potential to interface with WordNet.
1 INTRODUCTION
Language resources have a great impact on the
quality of any NLP application. The term “Language
resource” refers to any machine readable pieces of
information in any format providing information
about language being processed. For example,
language resources for a certain language can
provide the following information for a given word:
different senses with definitions and examples for
each sense, different word relations like synonyms
and antonyms, word morphological analysis and
Semantic information.
These information are important for many NLP
applications like word sense disambiguation, text
similarity, semantic search, text mining, opinion
mining, and many others. Fassieh (Attia et al., 2009)
is one example on these applications.
A lot of work is done in the field of language
resources for many languages like English, French,
German and many other European languages, but
very few is done to Arabic. Moreover, these few
Arabic language resources are limited and not fully
developed. Yaseen, et al. (2006) conducted a review
on Arabic language resources. A good language
resource can be done manually by expert linguists,
but such task can take a long time and too much
human labor to achieve.
In this paper we examine various Arabic
language resources, compare them and apply an
algorithm to integrate the scattered information
across these different resources into one compact
database using a configurable technique between
fully and semi-automated methods showing the
trade-off them in the integration.
The paper is organized as follows, section 2
discusses related work, while section 3 presents a
comparison of different Arabic resources. Section 4
presents the proposed integration methodology.
Section 5 presents the database architecture and
description. Section 6 presents the evaluation and
testing. Section 7 discusses the limitations and future
work, and finally we conclude our work in section 8.
2 RELATED WORK
Several attempts have been recorded to enrich the
Arabic language resources. Elkateb, et al. (2006)
reported on efforts for building a WordNet for
Arabic. They followed the methods developed for
EuroWordNet (Vossen, 1998). Concepts from
WordNet (Princeton University "About WordNet.",
233
Raafat H., Zahran M. and Rashwan M..
Arabase - A Database Combining Different Arabic Resources with Lexical and Semantic Information.
DOI: 10.5220/0004656702330240
In Proceedings of the International Conference on Knowledge Discovery and Information Retrieval and the International Conference on Knowledge
Management and Information Sharing (SSTM-2013), pages 233-240
ISBN: 978-989-8565-75-4
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
2010), EuroNet languages and BalkaNet (Tufis,
2004) are used as Synsets in ArabicWordNet. Then
some Arabic language specific concepts are
translated and added too. Equivalent English entries
according to SUMO ontology (Niles and Pease,
2001) are translated and added as well. Finally a bi-
directional propagation is performed from English to
Arabic and vice versa to generate Synsets. Most of
the work is done manually which decreased the
coverage and depth of the resulting resource.
Another attempt is bootstrapping an
ArabicWordNet using WordNet and parallel corpora
(Diab, 2004). She exploits the fact that a polysemous
word in one language will have a number of
translations in another language. These translations
can be clustered based on the word sense proximity
using WordNet. Diekema (2004) attempted to build
an English-Arabic semantic resource that can be
used in CLIR. She used WordNet and various
bilingual resources. Attia et al. (2008) built a rich
Arabic lexical semantic database based on the theory
of semantic field using various Arabic resources.
3 ARABIC LANGUAGE
RESOURCES
We started our work by searching for different
Arabic language resources, and exploring what
information they provide. First, we present the
nature of information provided by these resources,
and then we present a comparison between these
resources.
Below is a list of the information provided by
these Arabic language resources.
Morphological information (I
m
): The Word
morphological analysis, like root, type, gender, and
number.
Sense information (I
se
): Different word senses
with gloss, definitions and examples for each sense.
Synset information (I
sy
) (adapted from
WordNet
terminology): Different relations between
two sets of words. E.g. synonyms and antonyms.
Semantic information(I
sm
): The Semantic field
that the word belongs to together with different
semantic relations between semantic field pairs.
Interfacing with WordNet (I
wn
): Arabic words
are linked to their equivalent English words in
WordNet.
Each word can have one or more of these
information components in what we call the
‘information vector’ < I
m,
I
se,
I
sy,
I
sm,
I
wn
>
Next, we present the resources we found and we
give tabular comparison between them in table 2.
The main entry of all resources is the
unvocalized word which can have more than one
vocalized form. Vocalized forms are classified by
their part of speech (POS) into nouns, verbs and
particles. Each vocalized form can have its own
morphological information and it can have also more
than one sense. Each sense can have its sense
information, synset information, semantic
information and a WordNet interface if applicable.
3.1 King Abdulaziz City for Science
and Technology (KACST)
It is an Arabic-Arabic resource in a MYSQL
database format (almuajam, 2011). It provides sense
and morphological information. It has very limited
and incomplete semantic information (Figure1).
3.2 Arramooz
It is an Arabic-Arabic resource available in different
formats (SQL, XML and raw text) (Arramooz
AlWaseet, n.d.). It provides morphological and
sense information (Figure2).
3.3 Arabic WordNet (AWN)
It is an Arabic-English resource in derby database
format (Arabic WordNet, 2007). Its main entry is
both English and Arabic words. It provides Arabic-
Arabic information by mapping and translating the
pre-existing English-English information obtained
from WordNet, thus interfacing with it. It also
provides synset and sense information (Figure3).
3.4 RDI Lexical Semantic Database
It is an Arabic-Arabic resource by RDI available in
access format (Attia et al., 2008). It provides
semantic information and some sense,
morphological information (Figure4).
3.5 RDI Lite Lexicon
This resource is limited. It provides some
morphological information and some sense
information.
3.6 Alkhalil
It is an Arabic morphological analyser used to add
some morphological information when needed in the
integration (alkhalil dot net, 2011).
KDIR2013-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
234
3.7 Arabic Stop Words
It provides Arabic stop words with morphological
inflections using a combinations of possible prefixes
and suffixes (Arabic Stop Words, 2010). It has
10,389 unique entries.
Figure 1: Graphical representation for KACST.
Figure 2: Graphical representation for ARRAMOOZ.
Figure 3: Graphical representation for AWN.
Figure 4: Graphical representation for RDI-LSDB.
4 INTEGRATION
METHODOLOGY
The main goal of the integration is to try to allocate
an information vector from all available resources
for each distinct vocalized word. The integration
process is done in four main steps; Analysis, Design,
Integration, and Linking.
Arabase-ADatabaseCombiningDifferentArabicResourceswithLexicalandSemanticInformation
235
4.1 Analysis
Carefully analyse each resource to detect its
potential Information components. Then we
transform all the resources into one format to ease
the integration process. We transformed all the
resources to MYSQL database format.
4.2 Design
Put a design for the target integrated database given
the resources we have. The scalable design of
KACST database was the starting point of our
integrated database, then we modified it to match the
target database design as shown in figure 5. Also
table 2 show statistical comparison between the
existing and the target database, only the distinct and
non-floating entries are counted. Floating entries are
words entries with no information components.
4.3 Integration
Design and apply an algorithm to automatically
compile these resources together. The following is
an overview of the integration algorithm:
For each vocalized word w in resource r:
- Get (POS) of w if provided by r, otherwise use
Alkhalil to get POS of w
- Add w to a table corresponding to its POS only if
w was not added before, otherwise use the existing
entry.
- For each information i in r for w
- Add i to w
- For each vocalized word w
- Cluster information i for w by word sense.
‘i’ refers to an information component.
4.4 Linking
It is the information linking phase in which we
cluster the information content for each unvocalized
word by its senses.
5 ARABASE ARCHITECTURE
AND DESCRIPTION
5.1 Description
The main entry in the integrated resource is the
unvocalized word which has more than one
vocalized form. These vocalized forms can be
nouns, verbs, particles or unclassified. For example
consider the unvocalized word (w): (pronounced
Aien) it has more than one vocalized form
E.g. (w
1
):
, (w
2
):
Each of these vocalized forms can have more than
one sense. E.g. w
1
has these two senses.
Sense
1
(translated
as Eye). Sense
2
(translated as Spy).
Each sense can have a meaning or a definition,
and more than one example to illustrate its usage.
Each of these senses can have more than one
synonyms.
E.g. Sense
2
can have these two synonyms ,
. These synonyms form a Synset. We choose
one of these synonyms and promote it to be the
Synset head. We should note that synonyms
themselves are in fact vocalized words.
Each Synset can have a semantic field. E.g.
synset
1
{ , , } (Synonyms for the word
dark) can have the word (translated as color) as
its semantic field. Also the synset
2
{ , , }
(Synonyms for the word bright) can have as its
semantic field.
Synsets can have relations with each other E.g.
hyponym, antonym and synonym. E.g. synset
1
&
synset
2
are antonyms.
Semantic fields can have relationships with one
another as well.
Entries in the integrated resource are divided
among four main tables; Nouns, Verbs, Particles and
Not classified. The table ‘Not classified’ is for words
we couldn’t specify its (POS) during the integration
process because this information is missing from the
resource or Alkhalil failed to analyze the word.
Our goal is to compile as much information
vector components as possible for each vocalized
word w. These components will possibly be
compiled from different resources (r
1,
r
2 …
) so that
the final vector could be:
w :<r
1
(I
m
), r
2
(I
se
), r
3
(I
sy
), r
4
(I
sm
), r
5
(I
wn
)>. Where
r
x
(I
y
) is the piece of information y allocated from the
resource x.
5.2 Problems with Integration
Let w
1
and w
2
be two different senses from two
different resources for the same word w. Since
different senses of the same word can have the same
vocalized form of the word, therefore w
1
and w
2
have the same vocalized form. So we cannot rely
solely on the vocalized word form to distinguish
between different words senses, which means that
w
1
and w
2
could be the same sense or two different
senses. This confusion is a problem in the
KDIR2013-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
236
Figure 5: Graphical representation for Arabase.
integration. The following is an example on that
confusion:
For the vocalized word w
.
Resource r
1
provides <I
se,
I
m
>
w
1
:
I
se
:
(Definition for the word Eye).
I
m
: , , (Some morphological information).
Resource r
2
provides <I
se,
I
sm
>
w
2
:
I
se
:
(translated as Spy).
I
sm
: (Semantic field).
Resource r
3
provides <I
se,
I
sm
>
w
3
:
I
se
: (Definition for the
word Eye).
I
sm
: (Semantic field).
Resource r
4
provides <I
sy,
I
wn
>
w
4
:
I
sy
: , (Synonyms for the word Spy).
I
wn
: Spy (Corresponding WordNet entry).
We notice that w
1
, w
2,
w
3
and w
4
have the same
vocalized form but w
1
& w
3
have the same sense and
w
2
& w
4
have the same sense as well. Our goal is to
collect two information vectors one for each sense.
The first is for the sense (Eye)
< r
1
(I
m
), r
1
(I
se
) r
3
(I
se
), r
3
(I
sm
)> and the second is
for the sense (Spy) < r
2
(I
se
), r
2
(I
sm
), r
4
(I
sy
), r
4
(I
wn
)
>.
In order to do so, we have to search in the
available information for clues that infer that the
information from r
1
& r
3
and r
2
& r
4
belong
together. At the integration phase we assume that
each information vector I comes from resource r
represents a distinct sense. I.e. r
1
(I), r
2
(I), r
3
(I) and
r
4
(I) are totally different and each represents a
different sense (i.e. different synsets) for the word w
and finally at the linking phase we analyse these
information and link the related information
together.
5.3 Proposed Solutions
We discuss some heuristics we used in the linking
phase. The first: If two information components
belonging to two different resource are linked
together then all the information content of these
resources are linked together as well.
The second: All words in the same synset, i.e.
synonyms, share all the links established by all the
synset members. (Except for I
m
).
Generally all the sense information (I
se
) are
definitions, which means we can use text similarity
algorithms to decide if two given definitions are
similar or not, then using the heuristic discussed
above we link all the information content of their
resources.
If we look closely to the previous example we
find that we need to link r
1
& r
3
and r
2
& r
4
. We can
link r
1
(I
se
) with r
2
(I
se
) using text similarity
algorithms. Then we decide that r
1
& r
3
are linked
together to the same sense.
Generally, text similarity algorithms give a
similarity score for a given two texts, but it does not
decide if the two texts are similar or not. In this case,
to decide if two texts are similar or not using
similarity score algorithms we have the following
three alternatives.
The first is to use the similarity score when
querying the integrated database. We can show the
links with their scores or show only the links with
confidence exceed a certain value. It turned out that
this method can cause problems when using the
integrated database in different NLP applications.
The second is to find a threshold such that if the
similarity score between two texts is greater than
this threshold we label these two texts as similar and
not similar otherwise. The main drawback of this
method is the false positives (labelling two texts as
similar which in fact they are not) are dangerous
because the result of the text similarity decision is a
clue to link all the information of the two resources
these texts come from, which can cause serious
confusion.
The third is to do a Configurable semi-
automatic linking. There is no doubt that the linking
task and the confusion problem are solved using
Arabase-ADatabaseCombiningDifferentArabicResourceswithLexicalandSemanticInformation
237
human labor. But this is very time consuming task.
We propose to use text similarity techniques to
reduce the time taken by humans to do this task as
follows:
For a given two resources r
1
, r
2
having the
vocalized word w in common. We retrieve all the
senses for w per resources r
1
{sense
1
, sense
2
…}
,
r
2
{sense
1
, sense
2
…}
For each sense s in r
1
For each sense s` in r
2
s.calculateSimilarityScore(s`)
s.sortScoresDescendingly ()
This way each sense has a list of senses sorted
by similarity score. When one sense is chosen, all
the similar senses will appear sorted by similarity
score so there will be no need to scan all the senses
and most likely a match will be found in the first k-
best senses. The more accurate the text similarity
algorithm the less k will be before finding a match.
We compared two text similarity algorithms in terms
of the k-best measure; Modified Lesk and latent
semantic analysis.
5.3.1 Modified Lesk
Lesk (1986) method is used in word sense
disambiguation problems. We modified it to do text
similarity based on scoring two texts by counting the
number of common words between them. The
intuition behind that is that two texts expressing the
same meaning usually use similar words. We
modified the behaviour of the algorithm by
introducing some parameters to be tuned on a
development-set. Some of these parameters are
Boolean yes/no and others take real values. Boolean
parameters are: removing diacritization, removing
stop words, stemming the words and using edit
distance. Real valued ones: stop word to stop word
weight, stop word to non-stop word weight.
5.3.2 Latent Semantic Analysis
LSA (Deerwester et al., 1990) uses a corpus to build
a matrix whose rows represent unique words and
columns represent each paragraph (paragraphs in our
problem are the word’s definitions). Singular value
decomposition (SVD) is then used to reduce the
number of columns while preserving the similarity
structure among rows. Texts to be compared are
projected on this space and the similarity is
calculated based on the cosine of the angle between
the two vectors. The intuition behind using a
semantic similarity algorithm like LSA is that texts
expressing the same meaning are usually
semantically similar. The number of dimensions
used by LSA is yet to be tuned using a development-
set.
6 EVALUATION AND TESTING
We have to do two kinds of evaluations. The first is
the depth and breadth coverage evaluation of the
integrated resource itself. Breadth is the number of
entries found in the database, while depth is the
information content for each entry. The second is the
evaluation of the linking algorithm.
6.1 The Evaluation of the Integrated
Resource (Arabase)
In order to evaluate the database in both depth and
breadth coverage. We used a random sample of
running Arabic text collected form Wikipedia from
different topics, then for each word in the running
text we retrieved all the possible word forms from
Arabase (referred to as hits) this represents the
breadth coverage. For each hit we retrieved all its
possible information content (I
m,
I
se,
I
sy,
I
sm,
I
wn
) .This
represents the depth coverage. (Table 1).
Table 1: Depth and breadth coverage of Arabase.
6.2 The Evaluation of the Linking
Algorithm
We examined some words and manually link
information components together based on I
se
resulting in 184 links. We used 134 of them as a
development-set to tune the parameters for both
Lesk and LSA and we used the remaining 50 for
testing.
KDIR2013-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
238
Table 2: Shows a statistical comparison between the existing and the final integrated resources.
Figure 6: The performance of Lesk VS. LSA on the
development-set using the k-best measure.
Figure 7: The performance of Lesk VS. LSA on the test-
set using the k-best measure.
Figure 6 shows the performance of Lesk VS. LSA
on the development-set using the k-best measure.
The vertical axis is percentage of correct matches
sorted with a rank less than or equal to k. At k=10
both Lesk and LSA didn’t yet saturate and reach
100%. LSA saturates after k=11, while Lesk
saturates after k=17.
Figure 7 shows the performance of Lesk VS.
LSA on the test-set using the k-best measure. After
tuning both Lesk and LSA these parameters that
gave best performance on the development-set:
Lesk: Removing diacritization, edit distance: True.
Removing stop words, stemming: False
Stop word to stop word weight=0.01
Stop word to non-stop word weight=0.01
LSA: We used genism API (Rehurek, R., 2010) on a
445MB training corpus collected from (ksucorpus,
2013). The tuned number of dimensions=100.
7 LIMITATIONS AND FUTURE
WORK
Since the integration work here is automated with
adjustable human interaction, the algorithm is liable
to some errors that can be solved manually.
Below are some of the limitations involved in
our approach:
Classifying the words by POS is liable to errors
when using Alkhalil. If it failed to get the POS, the
word is given the label ‘Not classified.
N-gram entries (entries with more than one
token) are classified according to the POS
information given in the resource. If no such
information is found we classify it according to the
first token. If failed we label it as ‘Not classified’.
Poorly diacritized words can confuse the
morphological analyser, which ends up with more
than one morphological analysis for the same word.
50
55
60
65
70
75
80
85
90
95
100
12345678910
%ofcorrectmatches
k
Lesk LSA
50
55
60
65
70
75
80
85
90
95
100
1234567891011
%ofcorrectmatches
k
Lesk LSA
Arabase-ADatabaseCombiningDifferentArabicResourceswithLexicalandSemanticInformation
239
In these cases the first analysis is taken, which can
be erroneous approach but can be fixed manually.
We can also choose not to provide morphological
information for such words.
The linking algorithm is limited to linking
resources based on the sense information (I
se
). If any
resource does not have this information component
and its synset has no other words, then it will not be
linked by our linking algorithm.
We can enrich Arabase by linking it with
WordNet. Such that each Arabic sense is linked with
its corresponding English one in WordNet. Currently
the only interface is the integrated entries from
ArabicWordNet.
8 CONCLUSIONS
We compared different Arabic resources examining
their points of strength and weakness. Then we
presented a framework that can be used to compile
pieces of Arabic language information scattered
across these resources into a single resource. We
showed the trade-off between fully automated and
manual methods. Full automation will decrease
significantly the human effort, thus saving time and
man-power at the expense of decreasing the
accuracy and consistency of the resulting resources.
We showed the compromise between both methods
can result in an acceptable accuracy and consistency
with minimal human efforts.
REFERENCES
alkhalil dot net, 2011. KACST Available at:
http://sourceforge.net/projects/alkhalildotnet/.
[Accessed 23 June 2013].
almuajam, 2011. Arabic Interactive Dictionary Project.
Available at: http://sourceforge.net/projects/almuajam/
[Accessed 23 June 2013].
Arabic Stop Words, 2010. Available at:
http://arabicstopwords.sourceforge.net/. [Accessed 23
June 2013].
Arabic WordNet, 2007. A multi-lingual concept
dictionary, Available at: http://awnbrowser.
sourceforge.net/. [Accessed 23 June 2013].
Arramooz AlWaseet: Arabic dictionary for morphology.
Available at: http://arramooz.sourceforge.net/.
[Accessed 23 June 2013].
Attia, M., Rashwan, M., Al-Badrashiny M., 2009.
‘Fassieh; a Semi-Automatic Visual Interactive Tool
for the Morphological, PoS-Tags, Phonetic, and
Semantic Annotation of the Arabic Text’, IEEE
Transactions on Audio, Speech, and Language
Processing (TASLP): Special Issue on Processing
Morphologically Rich Languages.
Attia, M., Rashwan, M., Ragheb, A., Al-Badrashiny, M.,
Al-Basoumy, H. & Abdou, A., 2008. ‘A Compact
Arabic Lexical Semantics Language Resource Based
on the Theory of Semantic Fields’, Advances in
Natural Language Processing, LNCS vol. 5221, pp 65-
76.
Deerwester, S., Dumais, S.T., Furnas, G.W., Landauer,
T.K., & Harshman, R., 1990 ‘Indexing by Latent
Semantic Analysis’. Journal of the American society
for Information science, vol. 41, no. 6, pp. 391-407.
Diab, M., 2004. ‘The Feasibility of Bootstrapping an
Arabic WordNet leveraging Parallel Corpora and an
English WordNet’, Proceedings of the Arabic
Language Technologies and Resources, NEMLAR,
Cairo.
Diekema, A.R., 2004. ‘Preliminary Lexical Framework for
English-Arabic Semantic Resource Construction’,.
Semitic '04 Proceedings of the Workshop on
Computational Approaches to Arabic Script-based
Languages , Stroudsburg, PA, .pp. 10-14.
Elkateb, S., Black, W., Rodríguez, H., Alkhalifa, M.,
Vossen, P., Pease, A., & Fellbaum, C., 2006.
‘Building a WordNet for Arabic’, Proceedings of the
Fifth International Conference on Language
Resources and Evaluation (LREC 2006), Genoa, pp.
29-34.
ksucorpus, 2013. King Saud University Corpus of
Classical Arabic. Available at: http://
ksucorpus.ksu.edu.sa/ar/. [Accessed 23 June 2013].
Lesk, M., 1986. ‘Automatic sense disambiguation using
machine readable dictionaries: How to tell a pine cone
from an ice cream cone’, Proceedings of the 5th
annual international conference on Systems
documentation SIGDOC '86, Toronto, pp. 24-26.
Niles, I. & Pease, A., 2001. ‘Towards a standard upper
ontology’. Proceedings of the International
Conference on Formal Ontology in Information
Systems FOIS '01, Ogunquit, Maine, pp. 2-9.
Princeton University "About WordNet.” 2010.
WordNet.
Princeton University. Available at:
http://wordnet.princeton.edu. [Accessed 23 June
2013].
Rehurek, R. & Sojka, P., 2010. ‘Software Framework for
Topic Modelling with Large Corpora’, Proceedings of
the LREC 2010 Workshop on New Challenges for NLP
Frameworks, Valetta, pp. 46-50.
Tufis, D. (ed.), 2004. Special Issue on the BalkaNet
project. Romanian Journal of Information Science and
Technology, vol.7, no. 1-2
Vossen, P., 1998. ‘Introduction to EuroWordNet’,
Computers and the Humanities, vol. 32, no. 2-3, pp.
73-89.
Yaseen, M., Attia, M., Maegaard, B.... Rashwan, M., et.
al., 2006. ‘Building Annotated Written and Spoken
Arabic LR’s in NEMLAR Project’ Proceedings of the
Fifth International Conference on Language
Resources and Evaluation (LREC 2006), Genoa, pp.
533-538.
KDIR2013-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
240
