Automatic Annotation of Speech Events and Explicit Private State in Newswire
M. Arthur Munson · 2004
Expanding on preliminary results presented by Wiebe et al [18, 19], we investigate the feasibility of automatically annotating perspective and private state expressions in news text. A variety of supervised machine learning algorithms are used to identify single word annotations. The results reinforce the findings by Wiebe et al that the annotations are learnable. Further work is needed to include multi-word annotations in the learning framework and to use more informative learning features. 1