Building a Large Annotated Corpus of English: The Penn Treebank

4.3 Sample bracketing output

1. What did the early bracketing experiment find about annotation detail?
   Annotators were much faster correcting a simplified, skeletal structure; argument vs adjunct distinctions were hard and couldn’t be made consistently. So a skeletal representation without a forced argument/adjunct distinction was adopted.
   • Marcus et al., 1993, Building a Large Annotated Corpus of English (Computational Linguistics), §4.3
2. How is Fidditch’s output simplified before correction?
   POS tags, non-branching lexical nodes and some phrasal nodes (notably NBAR) are removed, flattening the structure; NP-internal structure up through the head is left unannotated. Illustration (not from the paper):

   ( (S (NP The annotators)
        (VP glued
            (NP the fragments)
            (PP into (NP one tree)))
        .))

   • Marcus et al., 1993, Building a Large Annotated Corpus of English (Computational Linguistics), §4.3
3. What are the X label and pseudo-attachment for?
   Letting annotators mark only structure they are sure of: X brackets a constituent of uncertain category; pseudo-attachment marks a constituent’s alternative attachment site, for “permanent predictable ambiguities” (where both attachments mean the same thing) and for the underlying position of extraposed elements.
   • Marcus et al., 1993, Building a Large Annotated Corpus of English (Computational Linguistics), §4.3
4. Why were ADJP brackets inside NPs left in the simplified structure for annotators to delete?
   Coordinated adjective phrases inside NPs must be bracketed, and deleting brackets with the mouse is much faster than creating them, so it was more efficient to leave ADJPs in and remove the unneeded ones.
   • Marcus et al., 1993, Building a Large Annotated Corpus of English (Computational Linguistics), §4.3