You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 8 Next »

Introduction

This guide will explain on how to do a labeling project with the Corpus Server, the Apache UIMA Cas Editor with the OpenNLP plugins.

Installing

Corpus Server

The Corpus server is running inside an OSGi Runtime Container. In this tutorial Apache Karaf will be used.

Install Apache Karaf

To install Apache Karaf follow the these steps:

Build the Corpus Server

Checkout the Corpus Server from svn with these two commands:

This will create a folder for each maven module.
Go first to the corpus-server, and do a mvn install there,
afterward do the same in the corpus-server-impl module.

Install the Corpus Server

Starting Apache Karaf via bin/karaf open the karaf console.
The following command will install all dependencies and the Corpus Server:

The corpus-server itself is just an interface layer which can expose an actual
Corpus Server implementation via its REST API to vairous tools which know this API.

To use the server an actual implementation must be installed in the OSGi runtime as
well, e.g the default implementation corpus-server-impl or a self made one.

This can be done with these commands:

The "list" command will now show the Corpus Server:
....
[  76] [Active     ] [            ] [   60] OpenNLP Corpus Server (0.0.1.SNAPSHOT)
[  77] [Active     ] [            ] [   60] Apache Derby 10.8 (10.8.1000002.1095077)
[  78] [Active     ] [            ] [   60] OpenNLP Corpus Server Implementation (0.0.1.SNAPSHOT)

Apache UIMA Cas Editor

Loading the Wikinews Corpus

The wikinews dumps can be downloaded from Wikipedia, here is some general information about the dumps:
http://meta.wikimedia.org/wiki/Data_dumps

The dump itself can be downloaded here: http://dumps.wikimedia.org/

Choose a mirror near you and then go to "Database backup dumps". The file you need to download is
called like this (the date will be different): enwikinews-20120727-pages-articles.xml.bz2

The current version of the parser only works well for the English wikinews dump. Contributions to fix this for other
languages are very welcome.

Checkout the wikinews parser:
svn co https://svn.apache.org/repos/asf/opennlp/sandbox/wikinews-importer/

And compile it with this command: mvn clean install

Opening an article in the Cas Editor

Setting up the OpenNLP annotator plugin

Training a component on the data

  • No labels