... than the baseline.Finally, we used our optimal configuration of TroFi, together with active learning and iterativeaugmentation, to build the TroFi Example Base, a publicly available, expandable ... deal of syntactic information in a single tag(each tag is an elementary tree from the XTAGEnglish Tree Adjoining Grammar). In additionto a word’s part of speech, they also encode in- formation ... UniversityBurnaby, BC, V 5A 1S6, Canadajbirke@alumni.sfu.ca, anoop@cs.sfu.caAbstract In this paper we present TroFi (TropeFinder), a system for automatically classi-fying literal and nonliteral usages of...