Paper: Features for Detecting Hedge Cues

ACL ID W10-3020
Title Features for Detecting Hedge Cues
Venue International Conference on Computational Natural Language Learning
Session Main Conference
Year 2010

We present a sequential labeling approach to hedge cue detection submitted to the bi- ological portion of task 1 for the CoNLL- 2010 shared task. Our main approach is as follows. We make use of partial syntac- tic information together with features ob- tained from the unlabeled corpus, and con- vert the task into one of sequential BIO- tagging. If a cue is found, a sentence is classified as uncertain and certain other- wise. To examine a large number of fea- turecombinations,weemployagenetical- gorithm. While some features obtained by this method are difficult to interpret, they wereshowntoimprovetheperformanceof the final system.