Abstract detail

Title Machine Translation of Idioms

In general, Machine Translation (MT) is a subfield of computational linguistics which investigates the use of computer software to translate text or speech from one natural language to another. We examine particularly the idiom processing within MT. A general definition of idiom is a figurative expression whose meaning is different from the simple sum of literal meanings of the individual words. We conduct experiments within the rule-based MT system CAT2. CAT2 is a unification- and transfer-based multilingual MT system that has been used since 1978 as an alternative to the EUROTRA software program. The system translates from and into several, not only west European languages, but also Russian, Chinese, Korean, Japanese, and Arabic. We wrote morphosyntactic rules, mainly for idiomatic verb phrases; an example follows: {lex=ksekayarizw,frame= {arg1={semf=pers}, arg2={nlu=logariasmos, agr={num=plu}}}}.[] <=> {lex=clear,frame= {arg1={semf=pers}, arg2={lex='the_air',agr={num=sing}}}}.[]. We “match” the Greek verb with the English one (ksekayarizw-clear), and the Greek noun with the English noun phrase (logariasmos-the air). Although, literally seen, there is no translation equivalence, this kind of matching leads to a successful translation result. Noteworthy is that in this specific idiom, the Greek noun, logariasmos, is in plural number, whereas its English correspondent, the air, is in singular.

Primary author
Dimitra Anastasiou