. // In: Dialogue'97: Computational Linguistics and its Applications. Proceedings of the Dialogue International Conference, 1997
Abstract The ETAP-3 English-to-Russian/Russian-to English MT system makes use of two main combinatorial dictionaries (ComD), the Russian ComD and the English ComD, both resorted to in each direction of translation. Recently, an important step has been made to introduce information on lexical functions (LF) of the Meaning Ы Text theory (MTT) into the ETAP-3 system. The information on LFs is considered to be independent of the translation purposes and is therefore located in the general zone of the ComDs. The notation used for representing LFs is close to that accepted in the standard MTT and, hence, quite familiar to NLP developers.
A major lexicographic problem to be solved when introducing LFs into MT is the determination of the space of LF values. Normally, an LF value is a lexeme; however, more complex values such as wordforms on the one hand and word combinations on the other hand are quite common in the MTT. In the present ETAP-3 release, every entered LF value is a lexeme which may however be followed by a preposition or a phrasal adverb. LFs are used at a few concrete points of the MT algorithm:
· After the syntactic structure of the sentence has been built, the parameter-type LFs are identified.
· In the transfer phase, the LF values are translated into the target language in accordance with the LF dictionary zone, overriding default equivalents.
· Subsequently, information on LFs is used to generate the missing prepositions and adverbs, wherever necessary.