Spoken translation system based on speech understanding in subject area. 
The automatic speech understanding system within a subject area is considered. 
The system firstly recognizes the input sentence as a sequence of words, 
than it makes decision about the meaning type of the recognized sentence.
Recognized words that missed in a target meaning type 
are caught by means of minimal edition distance estimation procedure. 
Strict, free and phonetic word based speech recognition word grammars for speech recognition are analysed. 
Results of their application for highly inflective languages with relatively free word order 
like Ukrainian are discussed and the respective translation system is presented.