Search code examples
nlpstanford-nlpstanford-parser

spaCy-like dependency graph navigation in CoreNLP

Is it possible to navigate the dependency parse tree in CoreNLP the way one does that in spaCy as described here? So far I see that token attributes like lemmas, POS tags, etc. are retrievable through an index, e.g. sent.lemmas(5) returns the lemma of the sixth token. I am not sure this exists for dependency heads and relations. Is there an established way of using these apart from navigating the whole tree every time?

Solution

  • If you're using the regular API, I believe what you're looking for is the function:

    Set<IndexedWord> SemanticGraph#vertexSet()

    This iterates over all of the nodes in a dependency tree [/graph]. each IndexedWord is also a CoreLabel, which means it has all of the functions you know and love for tokens.

    From the simple API -- which I gather is what you're using -- you can get a regular old dependency graph with:

    SemanticGraph Sentence#dependencyGraph()