measure(ment),1-4-E95-1021,ak </term> and compare the results . The <term> accuracy </term> of the <term> statistical method </term>
model,36-2-E95-1021,ak train and test the easy-to-implement <term> statistical model </term> . We describe the two <term> systems
other,14-4-E95-1021,ak comparable to <term> taggers </term> for <term> English </term> . But the <term> constraint-based tagger
other,17-5-E95-1021,ak limited time we allowed ourselves for <term> rule development </term> . We provide a unified account of
other,18-1-E95-1021,ak constraint-based disambiguation </term> , using <term> French </term> as our <term> test language </term> .
other,21-1-E95-1021,ak </term> , using <term> French </term> as our <term> test language </term> . We imposed a time limit on our
tech,12-1-E95-1021,ak <term> part-of-speech tagging </term> , <term> statistical and constraint-based disambiguation </term> , using <term> French </term> as our <term>
tech,12-4-E95-1021,ak is reasonably good , comparable to <term> taggers </term> for <term> English </term> . But the <term>
tech,19-2-E95-1021,ak of time spent on the design of our <term> constraint system </term> was about the same as the time we
tech,2-5-E95-1021,ak </term> for <term> English </term> . But the <term> constraint-based tagger </term> seems to be superior even with the
tech,4-3-E95-1021,ak model </term> . We describe the two <term> systems </term> and compare the results . The <term>
tech,4-4-E95-1021,ak results . The <term> accuracy </term> of the <term> statistical method </term> is reasonably good , comparable to
tech,9-1-E95-1021,ak compare two competing approaches to <term> part-of-speech tagging </term> , <term> statistical and constraint-based
隐藏细节