Run the NLP Pipeline

Contents

• NLP components.
• Command-ine tools.
• Using APIs.

NLP Components

ClearNLP supports NLP components for the following tasks.

• Tokenization and segmentation.
• Part-of-speech tagging.
• Morphological analysis.
• Dependency parsing.
• Semantic role labeling.

Command-Line Tools

• The following shows the command to run the NLP pipeline. The pos, morph, dep, srl modes perform the pipeline for part-of-speech tagging, morphological analysis, dependency parsing, and semantic role labeling, respectively.

  java com.clearnlp.nlp.engine.NLPDecode -z <mode> -c <filename> -i <filepath> [-ie <regex> -oe <string>]
  
  -z <mode>     : pos|morph|dep|srl
  -c <filename> : configuration file (required)
  -i <filepath> : input path (required)
  -ie <regex>   : input file extension (default: .*)
  -oe <string>  : output file extension (default: labeled)
  

• Download the following files and put them under clearnlp.

• config_decode.xml, log4j.properties, clearnlp.txt.

• If you are using the medical models, replace <model>general-en</model> with <model>medical-en</model>.

• The following command takes the input file clearnlp.txt, performs semantic role labeling (srl) using the configuration file config_decode.xml, and generates the output file clearnlp.txt.cnlp.

  $ java -Xmx4g -XX:+UseConcMarkSweepGC -Dlog4j.configuration=file:log4j.properties com.clearnlp.nlp.engine.NLPDecode -z srl -c config_decode.xml -i clearnlp.txt
  Loading feature templates.
  Loading models.
  Loading lexica.
  ...
  Decoding:
  /Users/jdchoi/Desktop/clearnlp.txt

• Use our visualization tool to view the output.

Using APIs with Maven and Eclipse

• Install maven: (guidelines).

• Create a maven project using the following command (use your own VERSION, GROUP_ID, and ARTIFACT_ID).

  VERSION=3.0
  GROUP_ID=edu.emory.clir
  ARTIFACT_ID=clearnlp-demo
  mvn archetype:generate -DarchetypeGroupId=org.apache.maven.archetypes -DgroupId=$GROUP_ID -DartifactId=$ARTIFACT_ID -Dversion=$VERSION-SNAPSHOT

• Add the following dependency to your pom.xml (here, we use the snapshot version).

  <dependency>
    <groupId>edu.emory.clir</groupId>
    <artifactId>clearnlp</artifactId>
    <version>3.0.0-SNAPSHOT</version>
  </dependency>
  

• Goto the clearnlp-demo directory and enter the following command.

  mvn eclipse:eclipse
  

• Create a Java project using the clearnlp-demo directory.