Extracting Synonyms from Bilingual Dictionaries Mustafa Jarrar Eman Karajah Birzeit University Birzeit University Palestine Palestine
[email protected] [email protected] Muhammad Khalifa Khaled Shaalan Cairo University The British University in Dubai Egypt United Arab Emirates
[email protected] [email protected] question answering, and machine translation Abstract among others. Synonyms are also considered essential parts in several types of lexical We present our progress in developing a resources, such as thesauri, wordnets (Miller et novel algorithm to extract synonyms from al., 1990), and linguistic ontologies (Jarrar, 2021; bilingual dictionaries. Identification and Jarrar, 2006). usage of synonyms play a significant role in improving the performance of There are different notions of synonymy in the information access applications. The idea literature varying from strict to lenient. In is to construct a translation graph from ontology engineering (see e.g., Jarrar, 2021), translation pairs, then to extract and synonymy is a formal equivalence relation (i.e., consolidate cyclic paths to form bilingual reflexive, symmetric, and transitive). Two terms are synonyms iff they have the exact same concept sets of synonyms. The initial evaluation of (i.e., refer, intentionally, to the same set of this algorithm illustrates promising results instances). Thus, T1 =Ci T2. In other words, given in extracting Arabic-English bilingual two terms T1 and T2 lexicalizing concepts C1 and synonyms. In the evaluation, we first C2, respectively, then T1 and T2 are considered to converted the synsets in the Arabic be synonyms iff C1 = C2. A less strict definition of WordNet into translation pairs (i.e., losing synonymy is used for constructing Wordnets, word-sense memberships).