Introduction

XBank is a simple tree-based merger of PropBank, NomBank, Discourse TreeBank and TimeBank annotations. Merger uses the Penn TreeBank phrase structure as the point of departure and finds a landing spot in the tree for each markable from PropBank, NomBank Discourse TreeBank and TimeBank. The landing spot is always a single token, usually the head of the markable (this may be changed in the future to allow entire consituents to be tagged).

The XBank browser is in essence a set of style sheets on top of the merged data. The merged files have not been changed except a few minor tweaks in order to attach the style sheet.

See the legend to see how tags from different schemes are represented. See statistics for some tag counts. Click on a document on the left to see it displayed in this frame. Click on "XBank Browser" on the top left to get this text back.

This is version 3 of the XBank Browser. Its main problem is that the merging script is real simplistic and that it merges in TimeBank tags at the wrong spot when the tokenization used for TimeBank is different from the one used for the TreeBank. This will be fixed in version 4.

Marc Verhagen, November 25th 2007.