Automated Alignment and Extraction of Bilingual Ontology for                   
Cross-Language Domain-Specific Applications  
Jui-Feng Yeh, Chung-Hsien Wu, Ming-Jun Chen and Liang-Chih Yu 
Department of Computer Science and Information Engineering 
National Cheng Kung University, Tainan, Taiwan, R.O.C. 
{jfyeh, chwu, mjchen,lcyu}@csie.ncku.edu.tw 
 
Abstract 
In this paper we propose a novel approach for 
ontology alignment and domain ontology extraction 
from the existing knowledge bases, WordNet and 
HowNet. These two knowledge bases are aligned to 
construct a bilingual ontology based on the co-
occurrence of the words in the sentence pairs of a 
parallel corpus. The bilingual ontology has the merit 
that it contains more structural and semantic 
information coverage from these two 
complementary knowledge bases. For domain-
specific applications, the domain specific ontology 
is further extracted from the bilingual ontology by 
the island-driven algorithm and the domain-specific 
corpus.  Finally, the domain-dependent 
terminologies and some axioms between domain 
terminologies are integrated into the ontology. For 
ontology evaluation, experiments were conducted 
by comparing the benchmark constructed by the 
ontology engineers or experts. The experimental 
results show that the proposed approach can extract 
an aligned bilingual domain-specific ontology. 
1 Introduction 
In recent years, considerable progress has been 
invested in developing the conceptual bases for 
building technology that allows knowledge reuse 
and sharing. As information exchangeability and 
communication becomes increasingly global, 
multilingual lexical resources that can provide 
transnational services are becoming increasingly 
important. On the other hand, multi-lingual 
ontology is very important for natural language 
processing, such as machine translation (MT), web 
mining (Oyama et al. 2004) and cross language 
information retrieval (CLIR). Generally, a multi-
lingual ontology maps the keyword set of one 
language to another language, or compute the co-
occurrence of the words among languages. In 
addition, a key merit for multilingual ontology is 
that it can increase the relation and structural 
information coverage by aligning two or more 
language-dependent ontologies with different 
semantic features.  
 Over the last few years, significant effort has 
been made to construct the ontology manually 
according to the domain expert’s knowledge. 
Manual ontology merging using conventional 
editing tools without intelligent support is difficult, 
labor intensive and error prone. Therefore, several 
systems and frameworks for supporting the 
knowledge engineer in the ontology merging task 
have recently been proposed (Noy and Musen 
2000). To avoid the reiteration in ontology 
construction, the algorithm of ontology merging 
(UMLS http://umlsks.nlm.nih.gov/) (Langkilde and 
Knight 1998) and ontology alignment (Vossen and 
Peters 1997) (Weigard and Hoppenbrouwers 1998) 
(Asanoma 2001) were invested. The final ontology 
is a merged version of the original ontologies. The 
two original ontologies persist, with aligned links 
between them. Alignment usually is performed 
when the ontologies cover domains that are 
complementary to each other. In the past, domain 
ontology was usually constructed manually 
according to the knowledge or experience of the 
experts or ontology engineers. Recently, automatic 
and semi-automatic methods have been developed. 
OntoExtract (Fensel et al. 2002) (Missikoff et al. 
2002) provided an ontology engineering chain to 
construct the domain ontology from WordNet and 
SemCor. 
Nowadays vast investment is made in ontology 
construction for domain application. Finding the 
authoritative evaluation for ontology is becoming a 
critical issue. Some evaluations are integrated into 
the ontology tools to detect and prevent the 
mistakes. The mistakes that might be made in 
developing taxonomies with frames are described in 
(Gómez-Pérez 2001). They defined three mainly 
types of mistakes: Inconsistency, Incompleteness, 
and redundancy. To deal with these mistakes and 
carry out the validation and verification of ontology, 
some ontology checkers, validators and parsers 
were developed. These tools provide the efficacious 
appraisal of correctness when developing the new 
ontology. However, they are disappointing in 
ontology integration, especial when the original 
ontologies are well defined. For other approaches 
(Maedche and Staab 2002), the similarity measures 
are proposed in the earlier stage of the evaluation. 
The evaluation consists two layers: lexical layer and 
conceptual layer. In lexical layer, the edit distance 
is integrated into the lexical similarity measure. The 
measure is defined as: 
()
()()
()
[]
min , ,
,max0, 0,1
min ,
ij ij
ij
ij
LL edLL
SM L L
LL
⎛⎞
−
⎜⎟
≡∈
⎝⎠
 (1) 
where 
( )SM null
 denotes the lexicon similarity 
function, ()ed null  is the Levensthein edit distance 
function defined in (Levensthein. 1966). 
i
L  and 
j
L  
are the words within the lexicons of the ontologies. 
The conceptual layer focuses on the conceptual 
structures of the ontologiesm namely taxonomic 
and nontaxonomic relations. 
In this paper, WordNet and HowNet knowledge 
bases are aligned to construct a bilingual universal 
ontology based on the co-occurrence of the words 
in a parallel corpus. For domain-specific 
applications, the medical domain ontology is further 
extracted from the universal ontology using the 
island-driven algorithm and a medical domain 
corpus. Finally, the axioms between medical 
terminologies are derived. The benchmark 
constructed by the ontology engineers and experts 
is introduced to evaluate the bilingual ontology 
constructed using the methods proposed in this 
paper. This paper defines two measures, taxonomic 
relation and non-taxonomic relation, as the 
quantitative metrics to evaluate the ontology.  
The rest of the paper is organized as follows. 
Section 2 describes ontology construction process 
and the web search system framework. Section 3 
presents the experimental results for the evaluation 
of our approach. Section 4 gives some concluding 
remarks. 
2 Methodologies 
Figure 1 shows the block diagram for ontology 
construction. There are two major processes in the 
proposed system: bilingual ontology alignment and 
domain ontology extraction. 
2.1 Bilingual Ontology Alignment 
In this approach, the cross-lingual ontology is 
constructed by aligning the words in WordNet to 
their corresponding words in HowNet. 
The hierarchical taxonomy is actually a 
conversion of HowNet. One of the important 
portions of HowNet is the methodology of defining 
the lexical entries. In HowNet, each lexical entry is 
defined as a combination of one or more primary 
features and a sequence of secondary features. The 
primary features indicate the entry’s category, 
namely, the relation: “is-a” which is in a 
hierarchical taxonomy. Based on the category, the 
secondary features make the entry’s sense more 
explicit, but they are non-taxonomic. Totally 1,521 
primary features are divided into 6 upper categories: 
Event, Entity, Attribute Value, Quantity, and 
Quantity Value. These primary features are 
organized into a hierarchical taxonomy. 
First, the Sinorama (Sinorama 2001) database is 
adopted as the bilingual language parallel corpus to 
compute the conditional probability of the words in 
WordNet, given the words in HowNet. Second, a 
bottom up algorithm is used for relation mapping. 
In WordNet a word may be associated with many 
synsets, each corresponding to a different sense of 
the word. For finding a relation between two 
different words, all the synsets associated with each 
word are considered (Fellbaum 1998). In HowNet, 
each word is composed of primary features and 
secondary features. The primary features indicate 
the word’s category. The purpose of this approach 
is to increase the relation and structural information 
coverage by aligning the above two language-
dependent ontologies, WordNet and HowNet, with 
their semantic features. 
 
 
Figure 1 Ontology construction framework 
The relation “is-a” defined in WordNet 
corresponds to the primary feature defined in 
HowNet. Equation (2) shows the mapping between 
the words in HowNet and the synsets in WordNet. 
Given a Chinese word, 
i
CW  , the probability of the 
word related to synset, 
k
synset  , can be obtained 
via its corresponding English synonyms, 
,  1,...,
k
j
EW j m=
 , which are the elements in 
k
synset  . The probability is estimated as follows.  
1
1
Pr( | )
Pr( , | )
(Pr( | , ) Pr( | ))
k
i
m
kk
ji
j
m
kk k
ji j i
j
synset CW
synset EW CW
synset EW CW EW CW
=
=
= ∑
=×∑
     (2) 
where 
()
()
Pr | ,
,,
,,
kk
ji
kk
jji
lk
jj i
l
synset EW CW
Nsynset EW CW
Nsynset EW CW
=
∑
     (3) 
In the above equation, 
( )
,,
kk
jji
Nsynset EW CW
  
represents the number of co-occurrences of 
i
CW  , 
k
j
EW
 and 
k
j
synset
. The probability 
()
Pr |
k
ji
EWCW
  is 
set to one when at least one of the primary features, 
( )
l
ii
PFCW
, of the Chinese word defined in the 
HowNet matches one of the ancestor nodes of 
synset , 
( )
k
j j
synset EW
 except the root nodes in the 
hierarchical structures of the noun and verb; 
Otherwise the probability 
( )
Pr |
k
ji
EWCW
  is set to zero. 
( )
()
( )
( )
Pr |
{,,,}
1 
((){,,,}
0
ji
l
ii
l
k
jj
k
EW CW
PF CW entity event act play
if
ancestor synset EW entity event act play
otherwise
⎧
−
⎪
⎪
=
⎨ −≠∅
⎪
⎪
⎩
UI
UU
                                                                               (4) 
where {enitity,event,act,play} is the concept set in 
the root nodes of HowNet and WordNet. 
Finally, the Chinese concept, 
i
CW
 , has been 
integrated into the synset , 
k
j
synset   , in WordNet 
as long as the probability, Pr
k
i
(synset |CW ) , is not 
zero. Figure 2(a) shows the concept tree generated 
by aligning WordNet and HowNet. 
2.2 Domain ontology extraction 
There are two phases to construct the domain 
ontology: 1) extract the ontology from the cross-
language ontology by the island-driven algorithm, 
and 2) integrate the terms and axioms defined in a 
medical encyclopaedia into the domain ontology.  
2.2.1 Extraction by island-driven algorithm 
Ontology provides consistent concepts and world 
representations necessary for clear communication 
within the knowledge domain. Even in domain-
specific applications, the number of words can be 
expected to be numerous. Synonym pruning is an 
effective alternative to word sense disambiguation. 
This paper proposes a corpus-based statistical 
approach to extracting the domain ontology. The 
steps are listed as follows: 
Step 1 Linearization: This step decomposes the 
tree structure in the universal ontology shown in 
Figure 2(a) into the vertex list that is an ordered 
node sequence starting at the leaf nodes and ending 
at the root node.  
Step 2 Concept extraction from the corpus: The 
node is defined as an operative node when the Tf-
idf value of word 
i
W   in the domain corpus is 
higher than that in its corresponding contrastive 
(out-of-domain) corpus. That is, 
_()
1,     ( ) ( )
0,   
i
Domain i Contrastive i
operative node W
if Tf idf W Tf idf W
Otherwise
−>−⎧
=
⎨
⎩
                                                                 (5) 
where 
,,
,
,
,,
,
,
()
log
()
log
Domain i
i Domain i Contrastive
iDomain
iDomain
Contrastive i
i Domain i Contrastive
i Contrastive
i Contrastive
Tf idf W
nn
freq
n
Tf idf W
nn
freq
n
−
+
=×
−
+
=×
 
In the above equations, 
Domaini
freq
,
  and 
eContrastivi
freq
,
  
are the frequencies of word 
i
W   in the domain 
documents and its contrastive (out-of-domain) 
documents, respectively. 
Domaini
n
,
  and  
,i Contrastive
n  
are the numbers of the documents containing word 
i
W   in the domain documents and its contrastive 
documents, respectively. The nodes with bold circle 
in Figure 2(a) represent the operative nodes.  
Step 3 Relation expansion using the island-
driven algorithm: There are some domain concepts 
not operative after the previous steps due to the 
problem of sparse data. From the observation in 
ontology construction, most of the inoperative 
concept nodes have operative hypernym nodes and 
hyponym nodes. Therefore, the island-driven 
algorithm is adopted to activate these inoperative 
concept nodes if their ancestors and descendants are 
all operative. The nodes with gray background 
shown in Figure 2(a) are the activated operative 
nodes.  
Step 4 Domain ontology extraction: The final 
step is to merge the linear vertex list sequence into 
a hierarchical tree. However, some noisy concepts 
not belonging to this domain ontology are operative. 
These nodes with inoperative noisy concepts should 
be filtered out. Finally, the domain ontology is 
extracted and the final result is shown in Figure 
2(b).  
After the above steps, a dummy node is added as 
the root node of the domain concept tree. 
 
 
Figure 2(a) Concept tree generated by aligning 
WordNet and HowNet. The nodes with bold circle 
represent the operative nodes after concept 
extraction. The nodes with gray background 
represent the operative nodes after relation 
expansion. 
 
Figure 2(b) The domain ontology after filtering out 
the isolated concepts 
2.2.2 Axiom and terminology integration 
In practice, specific domain terminologies and 
axioms should be derived and introduced into the 
ontology for domain-specific applications. There 
are two approaches to add the terminologies and 
axioms: the first one is manual editing by the 
ontology engineers, and the other is to obtain from 
the domain encyclopaedia.  
For medical domain, we obtain 1213 axioms 
derived from a medical encyclopaedia about the 
terminologies related to diseases, syndromes, and 
the clinic information. Figure 3 shows an example 
of the axiom. In this example, the disease 
“diabetes” is tagged as level “A” which represents 
that this disease is frequent in occurrence. And the 
degrees for the corresponding syndromes represent 
the causality between the disease and the 
syndromes. The axioms also provide two fields 
“department of the clinical care” and “the category 
of the disease” for medical information retrieval or 
other medical applications. 
 
 
Figure 3   One example of the axioms 
3 Evaluation 
For evaluation, a medical domain ontology is 
constructed. A medical web mining system is also 
implemented to evaluate the practicability of the 
bilingual ontology. 
3.1 Conceptual Evaluation of Ontology 
The benchmark ontologies are created to be the 
test-suites of reusable data which can be employed 
by ontology engineers or constructer for 
benchmarking purposes. The benchmark ontology 
was constructed by the domain experts including 
two doctors and one pharmacologist based on 
UMLS. The domain experts have integrated the 
Chinese concepts without changing the contents of 
UMLS  
Evaluation of ontology construction adopted the 
two layer measures: Lexical and Conceptual layers 
(Eichmann et al. 1998). The evaluation in the 
conceptual layer seems to be more important than 
that in the lexical layer when the ontology is 
constructed by aligning or merging several well 
defined source ontologies. There are two conceptual 
relation categories for evaluation: Taxonomic and 
non-Taxonomic evaluations. 
3.1.1 Evaluation of the taxonomic relation 
Step1 Linearization: This step decomposes the tree 
structure into the vertex list as described in Section 
2.2. The ontology,
T
O , and the benchmark, 
B
O are 
shown in the Figure 4(a) and 4(b), respectively. 
After this linearization, the vertex list sets: 
T
VLS and 
B
VLS  are obtained as shown in Figure 
4(c), where 
{ }
1234
,,,
TTTT
T
VLS VLVLVLVL=  
and 
{ }
123
,,
BBB
B
VLS VLVLVL= . 
 
(a) The taxonomic hierarchical representation of  
target ontology 
T
O  
 
(b) The taxonomic hierarchical representation of  
benchmark ontology 
B
O  
 
T
VLS  
B
VLS  
(c) The taxonomic vertex list set representation of  
target ontology and benchmark ontology 
Figure 4  Linearization of ontologies 
Step 2 Normalization: Since the frequencies of 
concepts in the vertex list set are not equal, the 
normalization factors are introduced to address this 
problem. For the target ontology, the factor vectors 
for normalization is 
{ }
12345678
,,,,,,,
T TTTTTTTT
NF nf nf nf nf nf nf nf nf= , 
and for the benchmark ontology is 
{ }
123456789
,,,,,,,,
B BBBBBBBBB
NF nfnfnfnfnfnfnfnfnf=
where 
o
i
nf   is the normalization factor for the i-th 
concept of the ontology O. It is defined as the 
reciprocal of the frequency in the vertex list set.   
i
1
the vertex lists contain the concept  in ontology O
O
i
nf =
 
Step 3 Estimation of the vertex list similarity: 
Therefore, the pairwise similarity of these two 
vertex lists of the target ontology and benchmark 
ontology can be obtained using the 
Needleman/Wunsch techniques shown in the 
following steps: 
Initialization: Create a matrix with m+1 columns 
and n+1 rows. m and n are the numbers of the 
concepts in the vertex lists of the target ontology 
and the bench mark ontology, respectively. The first 
row and first column of the matrix can be initially 
set to 0. That is, 
     (,) 0,  m 0  n 0 Sim m n if or= ==             (6) 
Matrix filling: Assign the values to the remnant 
elements in the matrix as the following equation: 
  
()
()
()
11 11
11
(, )
1
(1,1) (,),
2
1
max ( 1, )) ( , ),
2
1
(, 1) ( , )
2
ji
jjii
jjii
BT
mn
BBTT
mn lexiconmn
mn lexiconmn
BBTT
mn lexiconmn
Sim V V
Sim m n nf nf Sim V V
Sim m n nf nf Sim V V
Sim m n nf nf Sim V V
−− −−
−−
⎧
−−+ + ×
⎪
⎪
⎪
=−++×
⎨
⎪
⎪
−+ + ×
⎪
⎩
 
(7) 
There are some synonyms belonging to the same 
concept represented in one vertex. So the lexicon 
similarity can be described as   
1
1
1
(,)
Synonyms defined in the  and 
 Synonyms defined in the  or 
j
i
j
i
j
i
B
T
lexicon m n
B
T
mn
B
T
mn
Sim V V
VV
VV
−
−
−
=
  (8) 
Traceback: Determine the actual alignment with 
the maximum score, ,
j
i
B
T
mn
Sim(V V ), and therefore 
the pairwise similarity will be defined as the 
following equation: 
( )
,argmax ,
j
i
B
TTB
ij mn
Sim VL VL Sim(V V )≡  (9) 
Step 4 Pairwise similarity matrix estimation: 
The pairwise similarity matrix is obtained after 
p q× times for Step3. p ,q are the numbers of the 
vertex list of target ontology and benchmark 
ontology. Each element of the pairwise similarity 
matrix as Equation (10) is obtained using Equation 
(9).  
1
T
VL
T
p
VL
1
B
VL
B
q
VL
T
i
VL
B
j
VL
( )11,
TB
Sim VL VL
( ),
TB
p q
Sim VL VL
( ),
TB
ij
Sim VL VL
 
Figure 5 Pairwise similarity between the target 
ontolgy and benchmark ontology 
 
( )
() ()
() ()
11 1
1
,
, ... ,
::
, ... ,
TB
TB TB
q
TB TB
ppq
p q
PSM O O
Sim VL VL Sim VL VL
Sim VL VL Sim VL VL
×
⎡⎤
⎢⎥
≡
⎣⎦
O
           (10) 
Step 5 Evaluation of the taxonomic hierarchy: 
The whole similarity between target ontology and 
benchmark ontology can be represented as: 
( )
()
1
1
,
1
argmax ,
taxonomic T B
p
TB
ij
jq
i
Sim O O
Sim VL VL
p ≤≤
=
=
∑
            (11) 
3.1.2 Evaluation of the non-taxonomic relation 
Some relations defined in the ontology are non-
taxonomic set such as synonym. In fact, the lexicon 
similarity is applied to measure the conceptual 
similarity. The lexicon similarity of set can be 
defined as the following equation: 
(, )
Words defined in the  and 
 Words defined in the  or 
j
i
j
i
j
i
B
T
lexicon s t
B
T
st
B
T
st
Sim V V
VV
VV
=
          (12) 
Therefore, the evaluation of the non-taxonomic 
relation is defined as 
( )
11
,
1
(, )
j
i
non taxonomic T B
pq
B
T
lexicon s t
ij st
Sim O O
Sim V V
pq
−
==
=
×
∑∑∑∑
           (13) 
3.1.3 Evaluation Results 
Using the benchmark ontology and evaluation 
metrics described in previous sections, the 
evaluation results are shown in Table 1. 
Table1 the similarity measure between the target 
ontology and benchmark ontology 
Taxonomic relation similarity 0.57 
Non-Taxonomic relation similarity 0.68 
According to the experimental results, some 
phenomena are discovered as follows: first, the 
number of words mapped to the same concept in the 
upper layer of ontology is larger than that in the 
lower layer because the terminologies usually 
appear in the lower layer.  
3.2 Evaluation of domain application  
To assess the ontology performance, a medical 
web-mining system to search the desired page has 
been implemented. In this system the web pages 
were collected from several Websites and totally 
2322 web pages for medical domain and 8133 web 
pages for contrastive domain were collected. The 
training and test queries for training and evaluating 
the system performance were also collected. Forty 
users, who do not take part in the system 
development, were asked to provide a set of queries 
given the collected web pages. After post-
processing, the duplicate queries and the queries out 
of the medical domain are removed. Finally, 3207 
test queries using natural language were obtained. 
The baseline system is based on the Vector-Space 
Model (VSM) and synonym expansion. The 
conceptual relations and axioms defined in the 
medical ontology are integrated into the baseline as 
the ontology-based system. The result is shown in 
Table 2. The results show that ontology-based 
system outperforms the baseline system with 
synonym expansion, especially in recall rate. 
4 Conclusion 
A novel approach to automated ontology 
alignment and domain ontology extraction from 
two knowledge bases is presented in this paper. In 
this approach, a bilingual ontology is developed 
from two well established language-dependent 
knowledge bases, WordNet and HowNet according 
to the co-occurrence of the words in the parallel 
bilingual corpus. A domain-dependent ontology is 
further extracted from the universal ontology using 
the island-driven algorithm and a domain and its 
contrastive corpus. In addition, domain-specific 
terms and axioms are also added to the domain 
ontology. This paper also proposed an evaluation 
method, benchmark and metrics, for ontology 
construction. Besides, we also applied the domain-
specific ontology to the web page search in medical 
domain. The experimental results show that the 
proposed approach outperformed the synonym 
expansion approach. The overall performance of the 
information retrieval system is directly related to 
the ontology.  
 
 
Table 2 Precision rate (%) at the 11 points recall level 
Recall Level 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 
Baseline system 78 73 68 65 60 52 38 30 21 15 11 
Ontology based 87 86 82 77 73 71 68 62 51 40 32 

References  

N. Asanoma, 2001. Alignment of Ontologies: 
WordNet and Goi-Taikei. WordNet and Other 
Lexical Resources Workshop Program, 
NAACL2001. 89-94 

D. Eichmann, M. Ruiz, and P. Srinivasan, 1998. 
Cross-language information retrieval with the 
UMLS Metathesaurus, Proceeding of ACM 
Special Interest Group on Information Retreival 
(SIGIR), ACM Press, NY (1998), 72-80. 

D. Fensel, C. Bussler, Y. Ding, v. Kartseva1, M. 
Klein, M. Korotkiy, B. Omelayenko and R. 
Siebes, 2002. Semantic Web Application Areas, 
the 7th International Workshop on Applications 
of Natural Language to Information Systems 
(NLDB02). 

F. C. Fellbaum, 1998. WordNet an electronic 
Lexical Database, The MIT Press 1998. pp307-308 

A. Gómez-Pérez, 2001. Evaluating ontologies: 
Cases of Study IEEE Intelligent Systems and 
their Applications: Special Issue on Verification 
and Validation of ontologies. Vol. 16, Number 3. 
March 2001. Pags: 391-409. 

I. Langkilde and K. Knight, 1998. Generation that 
Exploits Corpus-Based Statistical Knowledge. In 
Proceedings of COLING-ACL 1998. 

V. Levensthein, 1966. Binary codes capable of 
correcting deletions, insertions, and reversals. 
Soviet Physics–Doklady, 10:707–710. 

A. Maedche, and S. Staab, 2002. Measuring 
Similarities between Ontologies. In Proceedings 
of the 13th European Conference on Knowledge 
Engineering and Knowledge Management 
EKAW, Madrid, Spain 2002/10/04 

M. Missikoff,, R. Navigli, and P. Velardi, 2002. 
Integrated approach to Web ontology learning 
and engineering, Computer, Volume: 35 Issue: 
11 . 60 –63 

N. F. Noy, and M. Musen, 2000. PROMPT: 
Algorithm and Tool for Automated Ontology 
Merging and Alignment, Proceedings of the 
National Conference on Artificial Intelligence. 
AAAI2000. 450-455 

S. Oyama, T. Kokubo, and T. Ishida, 2004. 
Domain-Specific Web Search with Keyword 
Spice. IEEE Transactions on Knowledge and 
Data Engineering, Vol 16,NO. 1, 17-27.  

Sinorama Magazine and Wordpedia.com Co., 2001. 
Multimedia CD-ROMs of Sinorama from 1976 to 
2000, Taipei. 
 
P. Vossen, and W. Peters, 1997. Multilingual 
design of EuroWordNet, Proceedings of the 
Delos workshop on Cross-language Information 
Retrieval. 

H. Weigard, and S. Hoppenbrouwers, 1998. 
Experiences with a multilingual ontology-based 
lexicon for news filtering, Proceedings in the 9th 
International Workshop on Database and Expert 
Systems Applications. 160-165 
