*****************************************************************************
*                                                                           *
*                                 README                                    *
*                                                                           *
*****************************************************************************

1. Instance Identifier

   Each instance in this dataset includes a unique identifier, formatted
   as follows:
   <sentence_id>__<"paired" word>_<index>__<"metaphor" word>_<index>

   Sentence IDs were determined based on where the instance's source
   sentence occurred in the original VUAMC dataset.  For instance, the
   first sentence in the dataset has an ID of 1, the 16th sentence in
   the dataset has an ID of 16, and so forth.  There are a total of 16202
   sentences in the original VUAMC.

   The "metaphor" word is the word in the pair that was originally labeled
   as a metaphor in the VUAMC.  In cases when both words were originally
   labeled as metaphors, it simply refers to the word for which potential
   pairings were being extracted at the time.

   The "paired" word is the other word in the pair (it may or may not have
   also been originally labeled as a metaphor).

   The index for each word corresponds to the token index for that word
   within the sentence.

2. Annotations from Trained Labelers

   Each test instance was annotated by two trained annotators who were
   native English speakers.  In cases in which the annotators only disagreed
   by a small amount (e.g., a 1 and a 2), their labels were averaged.  In
   cases of larger disagreement (e.g., a 0 and a 3), the annotations were
   adjudicated by a third party.

3. Contact

   If you have any questions about this dataset, please feel free to contact
   Natalie Parde at: natalie.parde@unt.edu.

Thanks!