SUBJECT: Re : &NAME Hi &NAME , I think it definitely makes sense for me to evaluate my thing on the &NAME . Subjects should be straightforward , I 'll read up on what &NAME did for objects and obliques . My algorithm is actually more involved than her baseline . for example , it looks like the base line would &NAME anything at &NUM position as a subject . mine would n't if it was in a &NAME or appositive . similarly , the basline would probably find an object relation between every verb and &NAME in &NUM position . mine would n't if there was a relation with the next verb , for example , in ' he said [ the ball ] was &NAME ' . Also her evaluation is for all &NAME . My evaluation is for only &NUM . So I do n't think my &NUM is incompatible with her baseline of &NUM . But it makes sense for me to evaluate against her as well for my thesis . the reason I have confidence in my algo is that I 've been through the mistakes my anaphora resolution algorithm makes and they tend not to be due to incorrect &NAME labelling . I actually did the evaluation on the gold standard afterwards , when I was writing the paper . I 'm not claiming that an approach so shallow can handle all &NAME , but I think it will do ok even on the &NAME for subjects and objects . - &CHAR .