================================================================================ LTAG-spinal Treebank contact: Libin Shen lshen@bbn.com Two versions of the LTAG-spinal treebank are provided in this package, v01 and v02. The difference between them is that, in v02, spinal templates and the representation of predicate coordination are normalized. For each version, the whole LTAG-spinal treebank is cut into five segments, according to the sections in the Penn Treebank, from with which the LTAG-spinal treebank is extracted. sec0-1: section 0-1 train: section 2-21 sec22: section 22 test: section 23 devel: section 24 There are 462 sentences in the Penn Treebank which are not included in the LTAG-spinal treebank. There section_IDs, file_IDs and sentence IDs are given in skip.txt. prop-all.idx provides a revised version of the Proposition bank. There are two changes 1. Word IDs are used to represent the phrases, while in the original Propbank, phrases are represented with the root node of the subtree in PTB. 2. Verb "be" is annotated in prop-all.idx, which is generated automatically. For copyright reasons, this public release does not include the mapping file to the Propbank annotations. Please contact us for a license. Please refer to the following website for more details of the treebank http://www.cis.upenn.edu/~xtag/spinal/ Please let me know if you have any suggestions or comments about the treebank. ================================================================================