other,4-1-P97-1058,ak </term> . Although adequate models of <term> human language </term> for <term> syntactic analysis </term>
tech,7-1-P97-1058,ak models of <term> human language </term> for <term> syntactic analysis </term> and <term> semantic interpretation </term>
tech,10-1-P97-1058,ak <term> syntactic analysis </term> and <term> semantic interpretation </term> are of at least <term> context-free
other,16-1-P97-1058,ak interpretation </term> are of at least <term> context-free complexity </term> , for applications such as <term> speech
tech,23-1-P97-1058,ak complexity </term> , for applications such as <term> speech processing </term> in which speed is important <term>
tech,30-1-P97-1058,ak </term> in which speed is important <term> finite-state models </term> are often preferred . These requirements
lr,10-2-P97-1058,ak reconciled by using the more complex <term> grammar </term> to automatically derive a finite-state
tech,27-2-P97-1058,ak then be used as a filter to guide <term> speech recognition </term> or to reject many hypotheses at an
tech,39-2-P97-1058,ak many hypotheses at an early stage of <term> processing </term> . A method is presented here for
lr,11-3-P97-1058,ak such finite-state approximations from <term> context-free grammars </term> . It is essentially different from
hide detail