The accompanying data sets are prepared by Peter LoBue (peter.lobue@temple.edu) of Temple University.

The RTE pairs are all taken from the RTE-5 corpus.

------------------------------------------------------------------------------------------------------

The accomanying data files Data_ALL.xlsx and Data_ALL.csv contain the authors' proofs to 108 select RTE-5 pairs. Each line in the proofs between the Text and Hypothesis indicates either a new piece of background knowledge brought to bear, or a modus ponens inference from the information in the text or previous lines of the proof.

Five annotators labeled 215 out of the 221 background knowledge statements with one of our 20 categories of world knowledge. All these results are also included.

Six statements were not categorized by annotators, either becasue we felt that they belonged in their own unique category or none of our categories at all. They are labeled in the data with an X.

Column A/1: RTE-5 pair ID, in the same row as the pair's Text
Column B/2: RTE-5 pair label (ENTAILMENT or CONTRADICTION), in the same row as the pair's Hypothesis
Column c/3: Text, individual proof statements, and Hypothesis of each RTE-5 pair
Column D/4: Most common category as labeled by annotators for background knowledge statements. Thirteen ties were broken by selecting from the top categories what the authors would have chosen.
Column E/5: Annotator #1's category selections
Column F/6: Annotator #2's category selections
Column G/7: Annotator #3's category selections
Column H/8: Annotator #4's category selections
Column I/9: Annotator #5's category selections