Determining Reliable Discourse Markers for Contrast Relations
Jennifer Spenader and Anna Lobanova
Eighth International Conference on Computational Semantics (IWCS-8 2009)
Tilburg University, Netherlands, January 7-9, 2009
Summary
Using simple statistics we studied to what degree there is evidence for the three-way distinction of Contrast relations, CONTRAST, ANTITHESIS and CONCESSION, recognized in standard Rhetorical Structure Theory (RST, Mann & Thompson 1987). Using the RST annotated corpus (Carlson et al. 2001) we show first that while the three Contrast relations share many discourse markers, a number of intuitive markers of Contrast, such as however, are actually not used statistically significantly more often than in Cause-Effect relations. We also present evidence that the three Contrast relations can be distinguished by which discourse markers they are statistically significantly more likely to occur. This result supports the three-way distinction and is counter evidence against discourse marking schemes that collapse the categories
START
Conference Manager (V2.56.8 - Rev. 414)