CoreNLP

The CoreNLP package is a thin Elixir client for the Stanford CoreNLP Server.

Since the Stanford offering is written in Java, the recommended integration for non-JVM languages is to stand up a CoreNLP dedicated server and hit it from a client interface. This package provides that interface for the Elixir ecosystem.

Installation

If available in Hex, the package can be installed by adding corenlp to your list of dependencies in mix.exs:

def deps do
  [{:corenlp, "~> 0.1.0"}]
end

The following are configuration settings used by this application to point to your CoreNLP server, with their defaults shown:

config :corenlp,
  host:          "localhost",
  base_path:     "/",
  port:          9000,
  recv_timeout:  30_000

Refer to the Stanford CoreNLP Server page for detailed information regarding standing up a dedicated server (including a local instance for dev/testing).

Usage

For detailed usage information regarding options that can be passed into the various calls, please consult the Stanford CoreNLP Server documentation.

Annotations

iex> CoreNLP.annotate("The cat sat.", annotators: "tokenize,ssplit,pos")
{:ok,
 %{"sentences" => [%{"index" => 0,
      "tokens" => [%{"after" => " ", "before" => "",
         "characterOffsetBegin" => 0, "characterOffsetEnd" => 3, "index" => 1,
         "originalText" => "The", "pos" => "DT", "word" => "The"},
       %{"after" => " ", "before" => " ", "characterOffsetBegin" => 4,
         "characterOffsetEnd" => 7, "index" => 2, "originalText" => "cat",
         "pos" => "NN", "word" => "cat"},
       %{"after" => "", "before" => " ", "characterOffsetBegin" => 8,
         "characterOffsetEnd" => 11, "index" => 3, "originalText" => "sat",
         "pos" => "VBD", "word" => "sat"},
       %{"after" => "", "before" => "", "characterOffsetBegin" => 11,
         "characterOffsetEnd" => 12, "index" => 4, "originalText" => ".",
         "pos" => ".", "word" => "."}]}]}}

TokensRegex

iex> CoreNLP.tokensregex("The quick brown fox jumps over the lazy dog.", ~S/(?$foxtype [{pos:JJ}]+ ) fox/)
{:ok,
 %{"sentences" => [%{"0" => %{"$foxtype" => %{"begin" => 1, "end" => 3,
          "text" => "quick brown"}, "begin" => 1, "end" => 4,
        "text" => "quick brown fox"}, "length" => 1}]}}

Semgrex

iex> CoreNLP.semgrex("The quick brown fox jumped over the lazy dog.", ~S|{pos:/VB.*/} >nsubj {}=subject >/nmod:.*/ {}=prep_phrase|)
{:ok,
 %{"sentences" => [%{"0" => %{"$prep_phrase" => %{"begin" => 8, "end" => 9,
          "text" => "dog"},
        "$subject" => %{"begin" => 3, "end" => 4, "text" => "fox"},
        "begin" => 4, "end" => 5, "text" => "jumped"}, "length" => 1}]}}

Tregex

iex> CoreNLP.tregex("The quick brown fox jumped over the lazy dog.", ~S|NP < NN=animal|)
{:ok,
 %{"sentences" => [%{"0" => %{"match" => "(NP (DT The) (JJ quick) (JJ brown) (NN fox))\n",
        "namedNodes" => [%{"animal" => "(NN fox)\n"}]},
      "1" => %{"match" => "(NP (DT the) (JJ lazy) (NN dog))\n",
        "namedNodes" => [%{"animal" => "(NN dog)\n"}]}}]}}

Contributing

All requests will be entertained, but the purpose of this package is to focus on providing services surrounding the Stanford CoreNLP Server specifically, and not to delve into the broader topic of NLP which will be left to other packages.

Documentation can be generated with ExDoc and published on HexDocs. Once published, the docs can be found at https://hexdocs.pm/corenlp.