Link
Search

Pipeline

Table of contents

Pipeline

The centerpiece of CoreNLP is the pipeline. Pipelines take in text or xml and generate full annotation objects.

Pipelines are constructed with Properties objects which provide specifications for what annotators to run and how to customize the annotators.

Running A Pipeline From The Command Line

You can immediately run a pipeline by issuing the following command:

java edu.stanford.nlp.pipeline.StanfordCoreNLP -file input.txt

The output will be stored in the file input.txt.out, and will by default contain a human readable presentation of the annotations.

You can customize your pipeline by providing properties in a properties file.

Here is an example properties file stored at example.props:

# output JSON instead
outputFormat = json

# list of annotators to run
annotators = tokenize,pos

# customize the pos model
pos.model = edu/stanford/nlp/models/pos-tagger/english-bidirectional-distsim.tagger

Here is the command for running a pipeline with these configurations:

java edu.stanford.nlp.pipeline.StanfordCoreNLP -props example.props -file input.txt

This will store JSON output in the file input.txt.json.

Of course all of those properties could be specified at the command line as well:

java edu.stanford.nlp.pipeline.StanfordCoreNLP -annotators tokenize,pos -pos.model edu/stanford/nlp/models/pos-tagger/english-bidirectional-distsim.tagger -outputFormat json -file input.txt

If you want to run a non-English language pipeline, you can just specify the name of one of the CoreNLP supported languages:

java edu.stanford.nlp.pipeline.StanfordCoreNLP -props french -file french-input.txt

Running A Pipeline In Java Code

Here is a basic demo class showing how to run a pipeline in Java code:

package edu.stanford.nlp.examples;

import edu.stanford.nlp.io.*;
import edu.stanford.nlp.pipeline.*;

import java.util.*;


public class PipelineExample {

    public static String text = "Marie was born in Paris.";

    public static void main(String[] args) {
        // set up pipeline properties
        Properties props = new Properties();
        // set the list of annotators to run
        props.setProperty("annotators", "tokenize,pos,lemma,ner,depparse");
        // build pipeline
        StanfordCoreNLP pipeline = new StanfordCoreNLP(props);
        // create a document object
        CoreDocument document = pipeline.processToCoreDocument(text);
    }

}

To customize pipelines in Java, add properties to the Properties object in the same way the annotators property is set in the code example.

Table of contents