Bibliome » AlvisNLP/ML

h1. Element Expression Examples

{{toc}}

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]].

h3. Sections in the corpus [corpus]

<pre>

documents.sections

</pre>

h3. Abstract sections [corpus]

<pre>

documents.sections:abstract

</pre>

or

<pre>

documents.sections(name == "abstract")

</pre>

First is faster.

h3. Abstract sections in the test set [corpus]

<pre>

documents(set == "test").sections:abstract

</pre>

This assumes that documents have a feature with key _set_ that denote the set to which it pertains.

h3. Documents whose PMID is in a file [corpus]

<pre>

documents(pmid in "good_pmids.txt")

</pre>

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).

h3. All genes and taxa [section]

<pre>

layer:genes | layer:taxa

</pre>

or

<pre>

layer('ne-tpe' == "gene" or 'ne-type' == "species")

</pre>

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.

h3. All words included in a sentence [annotation]

<pre>

inside:words

</pre>

This assumes that the context element is an annotation representing a sentence. It also assumes that all words are in a layer named _words_.

h3. All verbs [section]

<pre>

layer:words(pos ^= "V")

</pre>

or

<pre>

layer:words(pos =~ "^V")

</pre>

Both assume all word annotations are in a layer named _words_ and have a feature with key _pos_ whose value denote its POS. First should be slightly faster.

h3. All syntactic dependencies [section]

<pre>

relations:dependencies.tuples

</pre>

This assumes syntactic dependencies are in a relation named _dependencies_.

h3. Words that are subject [section]

<pre>

relations:dependencies.tuples(label == "SUBJ:V-N").args:dependent

</pre>

If you insist to check they are subjet to verbs:

<pre>

relations:dependencies.tuples(label == "SUBJ:V-N" and args:head.pos ^= "V").args:dependent

</pre>

h3. Subject of a verb [annotation]

<pre>

tuple:dependencies:head(label == "SUBJ:V-N").args:dependent

</pre>