lr,15-1-N03-1001,bq manual transcription </term> of <term> training data </term> . The method combines <term> domain
measure(ment),3-4-J05-4003,bq evaluate the <term> quality of the extracted data </term> by showing that it improves the performance
other,15-3-N03-4010,bq browsing the <term> repository </term> of <term> data objects </term> created by the <term> system
lr,13-4-P03-1033,bq learning </term> using real <term> dialogue data </term> collected by the <term> system </term>
lr,17-5-H05-1095,bq better generalize from the <term> training data </term> . This paper investigates some <term>
lr,27-3-H90-1060,bq the usual pooling of all the <term> speech data </term> from many <term> speakers </term> prior
lr,11-2-P03-1058,bq automatically acquire <term> sense-tagged training data </term> from <term> English-Chinese parallel
other,5-2-C92-1055,bq the problem of <term> insufficient training data </term> and <term> approximation error </term>
other,23-9-J05-1003,bq feature space </term> in the <term> parsing data </term> . Experiments show significant efficiency
other,13-1-P05-1067,bq statistical models </term> to <term> structured data </term> . In this paper , we present a <term>
</term> to recover a <term> submanifold </term> of data from a <term> high dimensionality space </term>
Understanding System ) </term> , which creates the data for a <term> text retrieval application </term>
lr,37-4-P03-1058,bq advantage that <term> manually sense-tagged data </term> have in their <term> sense coverage
lr,19-2-P01-1004,bq both <term> character - and word-segmented data </term> , in combination with a range of <term>
tech,9-1-H01-1049,bq paradigm for <term> human interaction with data sources </term> . We integrate a <term> spoken
lr,13-1-H05-1012,bq </term> based on <term> supervised training data </term> . We demonstrate that it is feasible
lr,9-1-H05-2007,bq <term> patterns </term> in <term> translation data </term> using <term> part-of-speech tag sequences
tech,5-3-P03-1058,bq this <term> method of acquiring sense-tagged data </term> is promising . On a subset of the
lr-prod,5-5-P06-1013,bq the <term> SemCor </term> and <term> Senseval-3 data sets </term> demonstrate that our ensembles
lr,43-2-N03-2003,bq bigger performance gains from the <term> data </term> by using <term> class-dependent interpolation
hide detail