Automatic Generation of Semantic Features and Lexical Relations Using owl ontologies

---

R. Muñoz et al. (Eds.): NLDB 2011, LNCS 6716, pp. 15–26, 2011. 
© Springer-Verlag Berlin Heidelberg 2011 
Automatic Generation of Semantic Features and Lexical 
Relations Using OWL Ontologies 
Maha Al-Yahya
1
, Hend Al-Khalifa
1
, Alia Bahanshal
2
, and Iman Al-Oudah
2
1
Information Technology Department, King Saud University, Riyadh, Saudi Arabia 
{malyahya,hendk}@ksu.edu.sa 
2
Research Institute, King Abdulaziz City for Science and Technology (KACST),
Riyadh, Saudi Arabia 
{abahanshal,ialoudah}@kacst.edu.sa 
Abstract. Semantic features are theoretical units of meaning-holding compo-
nents which are used for representing word meaning. These features play a vital 
role in determining the kind of lexical relation which exists between words in a 
language. Although such model of meaning representation has numerous
applications in various fields, the manual derivation of semantic features is a 
cumbersome and time consuming task. We aim to elevate this process by de-
veloping an automated semantic feature extraction system based on ontological 
models. Such an approach will provide explicit word meaning representation, 
and enable the computation of lexical relations such as synonym and antonymy. 
This paper describes the design and implementation of a prototype system 
used for automatically deriving componential formulae, and computing lexical 
relations between words from a given OWL ontology. The system has been 
tested on a number of ontologies, both English and Arabic. Results of the eval-
uation indicate that the system was able to provide necessary componential 
formulae for highly-axiomed ontologies. With regards to computing lexical re-
lations, the system performs better when predicting antonyms, with an average 
precision of 40%, and an average recall of 75%. We have also found a strong 
relation between ontology expressivity and system performance.