N07-1030 1000 . The features used for the anaphoricity classifier are quite simple . They include
N07-1030 paired t-tests ( p < .05 ) . The anaphoricity classifier has an average accuracy of 80.2
P04-1020 maps a given value of cr to the anaphoricity classifier trained with this cost ratio
P04-1020 recall will drop owing to the anaphoricity classifier 's misclassifications of truly
N07-1010 Cardie ( 2002a ) trains a separate anaphoricity classifier in addition to a coreference
N07-1030 use the following model for our anaphoricity classifier : Akfk ( i , ANAPH ) ) ( 2 )
P04-1020 attempts to resolve only NPs that the anaphoricity classifier determines to be anaphoric .
P04-1020 when RIPPER is used to train an anaphoricity classifier in the local approach , cr is
N07-1030 nonanaphoric ( by the imperfect anaphoricity classifier ) when in fact they were anaphoric
D09-1102 resolution systems . He used an anaphoricity classifier as a filter for coreference resolution
N07-1010 addition to a coreference model . The anaphoricity classifier is applied as a filter and only
P04-1020 ( For the purpose of training anaphoricity classifiers with different values of cr ,
N07-1030 contrast to systems which use the anaphoricity classifier as a filter for the coreference
N07-1030 coreference classifier with an anaphoricity classifier which acts as a filter during
N07-1030 as anaphors regardless of the anaphoricity classifier . This allows them to improve
N09-1065 to improving the output of an anaphoricity classifier both result in increased coreference
P04-1020 define the conservativeness of an anaphoricity classifier as follows . We say that classifier
P04-1020 training and presented to the anaphoricity classifier , which returns a value of ANAPHORIC
N09-1065 " improve " the output of the anaphoricity classifier by exploiting the dependency
P04-1020 For each data set , we train an anaphoricity classifier and a coreference classifier
hide detail