Welcome to conllu-path’s documentation!¶
Note
This project is under active development.
conllu-path is a python library for searching conllu trees. It was inspired by
the xpath XML query language, hence the name. It grew out of the need
(related to a PhD thesis) to search for various syntactic patterns in syntactically
annotated corpora. These were either Universal Dependencies
corpora or files generated by parsers trained on the above.
So this library targets conllu format files.
Searching for syntactic patterns programmatically using existing conllu libraries – for instance, detecting if a node is a verb and has a child that is a noun and that child has a child that is a preposition whose lemma is on:
>>> for node in sentence:
... if node.upos=='VERB':
... for child1 in node.children:
... if child1.upos=='NOUN':
... for child2 in child1.children:
... if child2.upos=='ADP' and child2.lemma=='on':
... report_something(node)
gets old very quickly. conllu-path replaces this kind of code with a search
that looks like this:
>>> for node in sentence.search('.//[upos=VERB]/[upos=NOUN]/[upos=ADP lemma=on]'):
... report_something(node)
The API and the description/specs of the search expression language are in the works. For now, check out the Usage section for a quick tutorial.