RTE-2 Preprocessed Datasets

We preprocessed the text and hypothesis of each pair in the development and test sets. The preprocessing includes sentence splitting using MXTERMINATOR (Reynar and Ratnaparkhi, 1997) and dependency parsing using MINIPAR (Lin, 1998).

Using the pre-processed data is optional, and it is allowed, of course, to use alternative tools for preprocessing. Note that since the preprocessing is done automatically it does contain some errors. We provide this data as-is, and give no warranty on the quality of the pre-processed data.
The preprocessed format for both text and hypothesis is the same:

• A list of sentences (only one for the hypothesis, but for some texts there are a couple of sentences).
• For each sentence, a list of nodes generated by MINIPAR. There are two node types: those that describe a word in the sentence and "artificial" nodes. They can be differentiated by their ID. Nodes describing a word contain a numeric id. Artificial nodes id start with 'E' and then a number.
• Each node that describes a word contains the word surface form, its lemma and part of speech tag.
• Each node that does not describe an actual word may contain a lemma and a part of speech tag.
• Each node, except the root node, contains its parent id and the dependency relation with its parent.

Download:
preprocessed development set
preprocessed test set

References:

Jeffrey C. Reynar and Adwait Ratnaparkhi. A Maximum Entropy Approach to Identifying Sentence Boundaries. In Proceedings of the Fifth Conference on Applied Natural Language Processing,March 31-April 3, 1997. Washington, D.C.

Dekang Lin. 1998. Dependency-based evaluation of MINIPAR. In Proceedings of the Workshop on Evaluation of Parsing Systems at LREC 1998, Granada, Spain.