A Language-Independent System for Generating Feature Structures from Interlingua Representations

Two main problems in natural language generation are lexical selection and syntactic structure determination. In interlingua approach to machine translation, determining sentence structures becomes more difficult, especially when the interlingua does not contain any syntactic information, in this paper, a knowledge-based computational model which handles these two problems in interlingua approach is presented. The developed system takes interlingua representations of individual sentences, performs lexical selection, and produces frame-based syntactic structures. The system takes all the information about the target language from knowledge resources, in other words its architecture is language-independent. The implemented system is tested with Turkish through small-sized resources such that its output can be fed into a previously developed tactical generator to produce the final realizations of Turkish sentences.