Negative filter
hybrid, generation, model 12
(384.2 per million)
tech,0-1-P95-1034,ak
improvement
</term>
over the original tests .
<term>
Large-scale natural language generation
</term>
requires the integration of vast
#27950The most sophisticated complex learning method offers a small, but statistically significant, improvement over the original tests.Large-scale natural language generation requires the integration of vast amounts of knowledge: lexical, grammatical, and conceptual.
tech,1-2-P95-1034,ak
, grammatical , and conceptual . A
<term>
robust generator
</term>
must be able to operate well even
#27971Arobust generator must be able to operate well even when pieces of knowledge are missing.
tech,9-4-P95-1034,ak
attack these problems , we have built a
<term>
hybrid generator
</term>
, in which gaps in
<term>
symbolic
#28007To attack these problems, we have built ahybrid generator, in which gaps in symbolic knowledge are filled by statistical methods.
other,6-3-P95-1034,ak
missing . It must also be robust against
<term>
incomplete or inaccurate inputs
</term>
. To attack these problems , we have
#27993It must also be robust againstincomplete or inaccurate inputs.
tech,21-4-P95-1034,ak
symbolic knowledge
</term>
are filled by
<term>
statistical methods
</term>
. We describe
<term>
algorithms
</term>
#28019To attack these problems, we have built a hybrid generator, in which gaps in symbolic knowledge are filled bystatistical methods.
tech,14-6-P95-1034,ak
</term>
can be used to simplify current
<term>
generators
</term>
and enhance their
<term>
portability
#28044We also discuss how the hybrid generation model can be used to simplify currentgenerators and enhance their portability, even when perfect knowledge is in principle obtainable.
tech,2-5-P95-1034,ak
statistical methods
</term>
. We describe
<term>
algorithms
</term>
and show experimental results . We
#28024We describealgorithms and show experimental results.
other,16-4-P95-1034,ak
generator
</term>
, in which gaps in
<term>
symbolic knowledge
</term>
are filled by
<term>
statistical methods
#28014To attack these problems, we have built a hybrid generator, in which gaps insymbolic knowledge are filled by statistical methods.
other,11-1-P95-1034,ak
the integration of vast amounts of
<term>
knowledge
</term>
: lexical , grammatical , and conceptual
#27961Large-scale natural language generation requires the integration of vast amounts ofknowledge: lexical, grammatical, and conceptual.
other,13-2-P95-1034,ak
to operate well even when pieces of
<term>
knowledge
</term>
are missing . It must also be robust
#27983A robust generator must be able to operate well even when pieces ofknowledge are missing.
other,23-6-P95-1034,ak
portability
</term>
, even when perfect
<term>
knowledge
</term>
is in principle obtainable . Aggregating
#28053We also discuss how the hybrid generation model can be used to simplify current generators and enhance their portability, even when perfectknowledge is in principle obtainable.
other,18-6-P95-1034,ak
generators
</term>
and enhance their
<term>
portability
</term>
, even when perfect
<term>
knowledge
#28048We also discuss how the hybrid generation model can be used to simplify current generators and enhance theirportability, even when perfect knowledge is in principle obtainable.