Need help with spark-corenlp?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

424 Stars 118 Forks GNU General Public License v3.0 54 Commits 19 Opened issues


Stanford CoreNLP wrapper for Apache Spark

Services available


Need anything else?

Contributors list

Stanford CoreNLP wrapper for Apache Spark

This package wraps Stanford CoreNLP annotators as Spark DataFrame functions following the simple APIs introduced in Stanford CoreNLP 3.7.0.

This package requires Java 8 and CoreNLP to run. Users must include CoreNLP model jars as dependencies to use language models.

All functions are defined under

  • cleanxml
    : Cleans XML tags in a document and returns the cleaned document.
  • tokenize
    : Tokenizes a sentence into words.
  • ssplit
    : Splits a document into sentences.
  • pos
    : Generates the part of speech tags of the sentence.
  • lemma
    : Generates the word lemmas of the sentence.
  • ner
    : Generates the named entity tags of the sentence.
  • depparse
    : Generates the semantic dependencies of the sentence and returns a flattened list of
    (source, sourceIndex, relation, target, targetIndex, weight)
    relation tuples.
  • coref
    : Generates the coref chains in the document and returns a list of
    (rep, mentions)
    chain tuples, where
    are in the format of
    (sentNum, startIndex, mention)
  • natlog
    : Generates the Natural Logic notion of polarity for each token in a sentence, returned as
    , or
  • openie
    : Generates a list of Open IE triples as flat
    (subject, relation, target, confidence)
  • sentiment
    : Measures the sentiment of an input sentence on a scale of 0 (strong negative) to 4 (strong positive).

Users can chain the functions to create pipeline, for example:

import org.apache.spark.sql.functions._
import com.databricks.spark.corenlp.functions._

val input = Seq( (1, "Stanford University is located in California. It is a great university.") ).toDF("id", "text")

val output = input .select(cleanxml('text).as('doc)) .select(explode(ssplit('doc)).as('sen)) .select('sen, tokenize('sen).as('words), ner('sen).as('nerTags), sentiment('sen).as('sentiment)) = false)

|sen                                           |words                                                 |nerTags                                           |sentiment|
|Stanford University is located in California .|[Stanford, University, is, located, in, California, .]|[ORGANIZATION, ORGANIZATION, O, O, O, LOCATION, O]|1        |
|It is a great university .                    |[It, is, a, great, university, .]                     |[O, O, O, O, O, O]                                |4        |


If you are a Databricks user, please follow the instructions in this example notebook.


Because CoreNLP depends on

3.x but Spark 2.4 depends on
2.x, we release
as an assembly jar that includes CoreNLP as well as its transitive dependencies, except
being shaded. This might cause issues if you have CoreNLP or its dependencies on the classpath.

To use

, you need one of the CoreNLP language models:
# Download one of the language models. 
# Run spark-shell 
spark-shell --packages databricks/spark-corenlp:0.4.0-spark_2.4-scala_2.11 --jars stanford-corenlp-3.9.1-models.jar


Many thanks to Jason Bolton from the Stanford NLP Group for API discussions.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.