Sample searches for PubAnnotation

KevinBretonnelCohen edited this page Nov 14, 2014 · 15 revisions

Problem that we are trying to help with

PropBank-style frame file:

VerbNet-style representation:

How can you find the linguistic patterns that should appear in the PropBank-style representation? How can you determine empirically what kinds of constraints should appear in the VerbNet-style representation?

Broad categories of searches:

  • Single annotation layer
  • Multiple annotation layers

Use cases:

  • In order to create a lexical resource, discover selectional restrictions on arguments of a predicate
  • In order to write a grammar, find examples of subcategorization frames

Single annotation layer searches

Arranged in order of complexity:

Word: find all sentences containing the word "binds"

Lemma: find all sentences containing any form of "bind"

Term: find all sentences containing a Sequence Ontology annotation

Syntax: find all compound nouns

Syntax: find all sentences or phrases containing a VP dominating two NPs

Multiple annotation layer searches

Word + term: find all sentences containing "binds" followed by a Sequence Ontology annotation

Example: NCBI Taxonomy within a noun phrase

Lemma + term: find all sentences containing "promotes" followed by any ontology concept

Syntax + term: find all VPs dominating any term

Syntax + term: find all VPs in which an argument of the verb is a term