P04-1017 expected to be also helpful for non-pronoun resolution . We would like to investigate
P03-1023 resolution , the improvement in non-pronoun resolution is slight . As shown in Table
P04-1017 non-pronoun is found by an additional non-pronoun resolution module . Our model can still
P04-1017 the capability of the existing non-pronoun resolution module . In our experiments we
P04-1017 depends on the performance of the non-pronoun resolution clas - sifier , DTnon −
J08-3002 curves of different models for non-pronoun resolution . the pronominal anaphora resolution
P04-1017 the low recall of the current non-pronoun resolution module . This should be owed
P03-1023 features are effective especially for non-pronoun resolution . 7 Conclusion In this paper
P07-1067 information was only applied to the non-pronoun resolution . For evaluation , Vilain et
J08-3002 section , here we find that for non-pronoun resolution the superiority of the twin-candidate
J08-3002 pronouns and non-pronouns . In non-pronoun resolution , an anaphor and its antecedent
P07-1067 with the pattern features for non-pronoun resolution ( NWire domain , filtered frequency
P04-1017 despite the low recall of the non-pronoun resolution module . In the current work
J08-3002 combining both pronoun resolution and non-pronoun resolution . Specifically , given an input
J08-3002 strongly indicative ( as with non-pronoun resolution ) or not ( as with pronoun resolution
P04-1017 Training Procedure : T1 . Train a non-pronoun resolution classifier DTnon − pron
P04-1017 two classifiers , one for the non-pronoun resolution and the other for pronoun resolution
J08-3002 0.8 % ) is larger than that for non-pronoun resolution ( 0.4 % , 1.4 % , and 0.6 % as
P03-1023 precision but a low recall in non-pronoun resolution ) . There - fore , most of the
P04-1017 pron is the classifier of the non-pronoun resolution module . One problem , however
hide detail