If you have reached this page, you are probably wondering what kind of patterns you can search for with CQP/Tree. In this page, we have gathered a few examples in depsearch and Grew-match. You can copy-paste them into the web interface and run them on SUK 1.1. With some adjustments, you can search for similar structures on any dependency-annotated corpus available on CLARIN.SI noSketch Engine, or even on other CQL-compatible corpus search systems.
Expletives are nominals that appear in an argument position of a predicate, but do not themselves fill any of the semantic roles of the predicate. This is a simple depsearch query to search for expletives attached to lexical verbs: {% with name="expl1-depsearch", code="VERB >expl _" %} {% include "snippet.html" %} {% endwith %}
or in Grew-match: {% with name="expl1-grew", code="""pattern { X -[expl]-> Y; X [ud_pos=VERB]; }""" %} {% include "snippet.html" %} {% endwith %}
These queries will fetch inherently reflective verbs, such as smejati se ('to laugh') but also other, non-reflexive expletives, such as piti ga ('to drink it'). If you are only interested in non-reflexive expletives, you can refine the query. In depsearch: {% with name="expl2-depsearch", code="VERB >expl PRON&!L=se" %} {% include "snippet.html" %} {% endwith %}
or in Grew-match: {% with name="expl2-grew", code="""pattern { X -[expl]-> Y; X [ud_pos=VERB]; Y [ud_pos=PRON, lemma <> \"se\"]; }""" %} {% include "snippet.html" %} {% endwith %}
As you can see in the examples above, one of the reasons why the depsearch queries are more concise is that some things are left implicit. In Grew-match, on the other hand, you always (excepts for dependency labels) have to specify the name of the annotation layers you want to search (in this case, lemma and ud_pos). This means that, if you want to run the Grew queries on a different corpus, you may have to modify these names. For Gigafida 2.2, for instance, ud_pos should be replaced with pos. To find out what different annotation layers are called in different corpora, select a corpus and click on . If your corpus of choice is not listed, select the corpus on noSketch Engine and click on "corpus info" to find out.
Let's look at a more complex depsearch query: {% with name="caserr-depsearch", code="VERB >obj NOUN&Case=Acc >aux|>advmod Polarity=Neg" %} {% include "snippet.html" %} {% endwith %}
This can be used to search for clauses with negative polarity with an object in accusative rather than genitive case, such as to ne vem ('I don't know this'), where to is incorrectly inflected for case.
The Grew-match equivalent is: {% with name="caserr-grew", code="""pattern { V [ud_pos = VERB]; N [ud_pos = NOUN, Case = Acc]; P [Polarity = Neg]; V -[obj]-> N; V -[aux|advmod]-> P; }""" %} {% include "snippet.html" %} {% endwith %}
So far, we have only looked at examples that rely exclusively on UD annotations. However, you can search for any other token-level attributes, as long as you know how the annotation layer is called for the corpus you want to use. As mentioned above, you can find out by selecting the corpus on noSketch Engine and clicking on "corpus info".
For instance, on SUK you can easily look for semantic actants (agents) which are syntactically not expressed as subjects: {% with name="agents-depsearch", code="srl_dep=\"ACT\"
In Grew-match: {% with name="agents-grew", code="""pattern { X -[^nsubj]-> Y; Y [srl_dep=\"ACT\"]; }""" %} {% include "snippet.html" %} {% endwith %}