Element expression examples
- Introduction
- [Sections in the corpus \corpus](#Sections-in-the-corpus-\corpus-1)
- [Abstract sections \corpus](#Abstract-sections-\corpus-1)
- [Abstract sections in the test set \corpus](#Abstract-sections-in-the-test-set-\corpus-1)
- [Documents whose PMID is in a file \corpus](#Documents-whose-PMID-is-in-a-file-\corpus-1)
- [All genes and taxa \section](#All-genes-and-taxa-\section-1)
- [All words included in a sentence \annotation](#All-words-included-in-a-sentence-\annotation-1)
- [All verbs \section](#All-verbs-\section-1)
- [All syntactic dependencies \section](#All-syntactic-dependencies-\section-1)
- [Words that are subject \section](#Words-that-are-subject-\section-1)
- [Subject of a verb \annotation](#Subject-of-a-verb-\annotation-1)
The type of the expected kind of context element is given between brackets in the example title. For detailed information on all the expression parts, you may refer to Element expression reference.
documents.sections
documents.sections:abstract
or
documents.sections[@name == "abstract"]
documents[@set == "test"].sections:abstract
This assumes that documents have a feature with key set that denote the set to which it pertains.
documents[@pmid in "good_pmids.txt"]
This assumes that the PMID of the document is in a feature with key pmid, and there is a file in the current directory named good_pmids.txt containing all PMIDs of interest (one per line).
layer:genes | layer:taxa
or
layer[@'ne-tpe' == "gene" or @'ne-type' == "species"]
The first is faster. It assumes that all gene annotations are in a layer named genes, and that all taxon annotations are in a layer taxa. The annotations are given in the following order: first genes in standard order, then taxa in standard order.
The second assumes that annotations have a feature names ne-type containing the named entity type of the annotation. The annotations are given in standard order regardless of the type.
inside:words
This assumes that the context element is an annotation representing a sentence. It also assumes that all words are in a layer named words.
layer:words[@pos ^= "V"]
or
layer:words[@pos =~ "^V"]
Both assume all word annotations are in a layer named words and have a feature with key pos whose value denote its POS.
relations:dependencies.tuples
This assumes syntactic dependencies are in a relation named dependencies.
relations:dependencies.tuples[@label == "SUBJ:V-N"].args:dependent
If you insist to check they are subjet to verbs:
relations:dependencies.tuples[@label == "SUBJ:V-N" and args:head.@pos ^= "V"].args:dependent
tuple:dependencies:head[@label == "SUBJ:V-N"].args:dependent