SCIE Classifiers

SCIE - Spinal Cord Injury Ontology Extraction
Copyright (C) 2013, 2014
Raphael Dickfelder, Jan Göpfert, Benjamin Paaßen, Andreas Stöckel

This program is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

Contents

  1. Introduction
  2. Installation
  3. Quickstart Guide
  4. Preprocessing
  5. Creating your Training Data
  6. Classifier Architecture
  7. Reading your Training Data
  8. Training your classifiers
  9. Creating Relation Annotations
  10. Dependencies
  11. GNU AFFERO GENERAL PUBLIC LICENSE

Introduction

This software has the purpose to construct relations in a pre-processed text relying on the results of other UIMA annotators. This approach is not domain-specific, but you need to know fairly well what you are doing to use this software productively.

It is only recommended to use this software if:

Installation

It is recommended to use this project with maven. To do that you should install Java 1.7 (or higher), install maven and then run the command

mvn install

Then the software is installed in your local maven repository and can be declared as dependency like so:

<dependency>
	<groupId>de.unibi.techfak.scie</groupId>
	<artifactId>Classifiers</artifactId>
	<version>1.1</version>
	<scope>compile</scope>
</dependency>

If you do not want to use maven you can build the sources to a .jar file using the command

mvn package

Quickstart Guide

It is strongly recommended to read this whole README. If you want to skip it you should at least copy the example at

src/test/java/de/unibi/techfak/scie/classifiers/RelationExtractionTest.java

and start your own implementation from there.

Preprocessing

For preproessing of your data you need some kind of Named Entitiy Recognition. It is possible to use the SCIE NER, but you can also plug in your own mechanism. Note that your named entities have to be annotated as UIMA classes in a JCas instance. The only necessary pre-requisite are tokens in the text, because many of the features used later depend on text tokenization. An example to do tokenization is the Stanford POS Tagger, which can be implemented as a UIMA annotator like this:

/*
 * SCIE -- Spinal Cord Injury Ontology Extraction
 * Copyright (C) 2013, 2014
 * Raphael Dickfelder, Jan Göpfert, Benjamin Paaßen, Andreas Stöckel
 * 
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Affero General Public License as
 * published by the Free Software Foundation, either version 3 of the
 * License, or (at your option) any later version.
 * 
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Affero General Public License for more details.
 * 
 * You should have received a copy of the GNU Affero General Public License
 * along with this program.  If not, see .
 */

package de.unibi.techfak.isy.lppba.annotators.structure;

import de.unibi.techfak.isy.lppba.descriptors.Sentence;
import de.unibi.techfak.isy.lppba.descriptors.Token;
import edu.stanford.nlp.ling.HasWord;
import edu.stanford.nlp.ling.TaggedWord;
import edu.stanford.nlp.tagger.maxent.MaxentTagger;
import java.io.StringReader;
import java.util.ArrayList;
import java.util.List;
import org.apache.uima.analysis_engine.AnalysisEngineProcessException;
import org.apache.uima.jcas.JCas;

/**
 * This uses the Stanford POS Tagger to tokenize sentences and words and give the words a POS
 * feature.
 *
 * The code is derived from the TaggerDemo given by Stanford itself.
 * 
 * The part of speech tags given by the Stanford tagger are described here:
 * 
 * http://www.ling.upenn.edu/courses/Fall_2003/ling001/penn_treebank_pos.html
 *
 * @author Benjamin Paassen - bpaassen(at)techfak.uni-bielefeld.de
 */
public class StanfordTokenizer extends
	org.apache.uima.fit.component.JCasAnnotator_ImplBase {

	private static final MaxentTagger tagger = new MaxentTagger(
		"importdata/english-left3words-distsim.tagger");

	public StanfordTokenizer() {
	}

	@Override
	public void process(JCas jcas) throws AnalysisEngineProcessException {
		final List<List<HasWord>> sentences
			= MaxentTagger.tokenizeText(new StringReader(jcas.getDocumentText()));
		for (final List<HasWord> sentence : sentences) {
			final ArrayList<TaggedWord> tSentence = tagger.tagSentence(sentence);
			if (!tSentence.isEmpty()) {
				//annotate a sentence with first and last word.
				final int sentenceStart = tSentence.get(0).beginPosition();
				final int sentenceEnd = tSentence.get(tSentence.size() - 1).endPosition();
				final Sentence uimaSentence = new Sentence(jcas, sentenceStart, sentenceEnd);
				uimaSentence.addToIndexes();
				for(final TaggedWord tWord : tSentence){
					final Token uimaWord = new Token(jcas, tWord.beginPosition(), tWord.endPosition());
					uimaWord.setPOS(tWord.tag());
					uimaWord.addToIndexes();
				}
			}
		}
	}
}

Note that the

Token
and the
Sentence
class are used from the SCIE type system. This is an important fact. Both classes are defined in the SCIESyntax.xml type system, which is provided along with the source code.

It is also necessary that your annotations extend the

de.unibi.techfak.isy.lppba.descriptors.Annotation
class, which is defined in the SCIEBase.xml. That only means that you need to add a unique ID to each instance that shall be used for relations later on. The IDs are required for training data generation (see below).

For our "IsIn"-Relation example a type system could look like this:


<!--
 * SCIE - Spinal Cord Injury Ontology Extraction
 * Copyright (C) 2013, 2014
 * Raphael Dickfelder, Jan Göpfert, Benjamin Paaßen, Andreas Stöckel
 * 
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Affero General Public License as
 * published by the Free Software Foundation, either version 3 of the
 * License, or (at your option) any later version.
 * 
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Affero General Public License for more details.
 * 
 * You should have received a copy of the GNU Affero General Public License
 * along with this program.  If not, see http://www.gnu.org/licenses/.
-->

<typeSystemDescription  xmlns="http://uima.apache.org/resourceSpecifier">
	<name>TrainExampleTypeSystem</name>
	<description>An example type system for the Relation Extraction Tutorial.</description>
	<vendor>Raphael Dickfelder, Jan Göpfert, Benjamin Paassen, Andreas Stoeckel</vendor>
	<version>1.0</version>
	<imports>
		<import location="SCIEClassifiers.xml"/>
	</imports>
	<types>
		<typeDescription>
			<name>de.unibi.techfak.scie.descriptors.City</name>
			<description>Annotates a city.</description>
			<supertypeName>de.unibi.techfak.scie.descriptors.Annotation</supertypeName>
		</typeDescription>
		<typeDescription>
			<name>de.unibi.techfak.scie.descriptors.Country</name>
			<description>Annotates a country.</description>
			<supertypeName>de.unibi.techfak.scie.descriptors.Annotation</supertypeName>
		</typeDescription>
		<typeDescription>
			<name>de.unibi.techfak.scie.descriptors.IsIn</name>
			<description>Annotates an "isIn"-relation.</description>
			<supertypeName>de.unibi.techfak.scie.descriptors.Aggregated</supertypeName>
			<features>
				<featureDescription>
					<name>subject</name>
					<description>The city that is in the object.</description>
					<rangeTypeName>de.unibi.techfak.scie.descriptors.City</rangeTypeName>
				</featureDescription>
				<featureDescription>
					<name>object</name>
					<description>The country the subject is in.</description>
					<rangeTypeName>de.unibi.techfak.scie.descriptors.Country</rangeTypeName>
				</featureDescription>
			</features>
		</typeDescription>
	</types>
</typeSystemDescription>

This type system can also be found at

src/main/resources/de/unibi/techfak/scie/descriptors/TrainExampleTypeSystem.xml

Creating your Training Data

As the name suggests this relation extraction mechanism is based on classifiers. To minimize your effort in producing training data you only need to provide positive training examples. The negative ones will be constructed automatically.

Training data consists of tuples of XCAS-files that contain your pre-processed data including all detected named entities and a .rel file that defines which named entities within that XCAS file should be aggregated to relations.

The syntax for .rel files is very easy. Here is an example:

ISIN:1,5
ISIN:2,6
ISIN:3,7

Each line consists of the name of the relation, then a colon and finally a comma-separated list of ids, which refer to entities in the XCAS file. That is the reason why you must generate the unique IDs beforehand. Note that the IDs must only be unique within one XCAS file. They do not need to be unique globally.

Also note that the .rel files must usually be written manually. Our experience shows that it is most comfortable to run your Named Entity generation on some of your input text files, then look on the XCAS output using the Apache UIMA Annotation Viewer (which is part of the UIMA SDK) and write the .rel-file manually based on that.

Classifier Architecture

Let's take a step back for a moment and consider again, what we are doing: We have defined relations as templates, that are to be filled by named entity instances of the correct type. Take the easy example of the "IsIn"-relation and the example text:

The Trans-Mongolian Railway connects Ulan-Ude, on the Trans- Baikal (Trans-Siberian) railway in Russia, with the Chinese city of Jining, by way of Ulaanbaatar in Mongolia.

which is taken from the english Wikipedia page on the Trans-Mongolian Railway. Assume we have preprocessed the text correctly and have detected the cities "Ulan-Ude", "Jining" and "Ulaanbaatar" as well as the country references "Mongolian", "Russia", "Chinese" and "Mongolia".

Now the set of possible "IsIn"-instances is the cartesian product of the set of city instances and the set of country instances, meaning

The naive approach would be to explicitly construct all these possible relation instances and then apply a binary classifier to decide whether that relation is correct or not. But the set of possible relation instances gets exponantially large in the number of slots. In case of a five-slot relation like "Animal" this can get infeasible very fast.

Fortunately we can exclude a lot of possible relations beforehand and do not need the full information of the possible relation for that. We can limit the number of possible candidates for each slot beforehand and thus restrict the number of possible combinations. Therefore we do not consider only a final relation classifier but three types of binary classifiers:

That distinction between different types of classifiers will become relevant in the rest of this README.

Reading your Training Data

For each of your classifiers you have to provide an implementation of the

TrainingDataReader
Interface that puts out
LabeledDataPoint
instances. However you do not need to start from scratch. For each of the three types mentioned above we provide an abstract implementation that you can extend.

CoreDataPoint

For trigger/core classifiers you should extend the

CoreDataPoint
class and the nested
CoreTrainingDataReader
. For our "IsIn"-example, where we regard a country instance as trigger, the implementations could look like this.

private static class CountryDataPoint extends CoreDataPoint<Country> {

	public CountryDataPoint(Country core, JCas jcas) {
		super(core, jcas);
	}

	@Override
	public void addSpecialFeatures(FeatureMap stdFeatures, boolean addFeaturesToDictionary) {
		//we have no special features here.
	}

	private static class TrainingDataReader extends CoreTrainingDataReader<Country> {

		public TrainingDataReader() {
			//Here we should specify the core class and the
			//name of the relation we used in the .rel file.
			super(Country.class, "ISIN");
		}

		@Override
		public CoreDataPoint<Country> createDataPoint(Country core, JCas jcas) {
			return new CountryDataPoint(core, jcas);
		}
	}
}

Note that the implementation is fairly straightforward. The constructor is trivial, the

addSpecialFeatures
method can be left empty in most cases (but you can provide your own features for classifier training, if you wish), the
TrainingDataReader
constructor specifies the class of our core as well as an arbitrary name for the relation and the
createDataPoint
method just has to call the
CountryDataPoint
constructor.

CoreSlotCombinationDataPoint

For slot classifiers you should extend the

CoreSlotCombinationDataPoint
class and the nested
CoreSlotTrainingDataReader
. For our "IsIn"-example, where we regard a city instance as slot, the implementations could look like this.

private static class CityDataPoint extends CoreSlotCombinationDataPoint<Country, City> {

	public CityDataPoint(Country core, City slot, JCas jcas) {
		//here we just want to specify the sigma parameter. Otherwise we
		//can redirect everything to the constructor of the super class.
		super(core, slot, jcas, 6);
	}

	@Override
	public void addSpecialFeatures(FeatureMap stdFeatures, boolean addFeaturesToDictionary) {
		//we have no special features here.
	}

	private static class TrainingDataReader extends CoreSlotTrainingDataReader<Country, City> {

		public TrainingDataReader() {
			//Here we should specify the core class, the slot class and the
			//name of the relation we used in the .rel file.
			super(Country.class, City.class, "ISIN");
		}

		@Override
		public CoreSlotCombinationDataPoint<Country, City>
				createCombination(Country core, City slot, JCas jcas) {
			return new CityDataPoint(core, slot, jcas);
		}
	}
}

The implementation here basically functions the same way as before. The only specialty is the sigma parameter in the

CoreSlotCombinationDataPoint
constructor. It can be interpreted as the distance (in words), you would typically expect between a country and a city instance that correctly belong together. This will be used as the standard deviation of a gaussian kernel on the distance between core and slot instance, which in turn is a feature for classification. The parameter is not that important. In the worst case it won't help the classifier. We used values between 6 and 15 in our application.

RelationDataPoint

RelationDataPoints function basically the same as before, but the TrainingDataReader is notably more important. The first difference is its constructor, where you have to specify the trigger classifier and all slot classifiers that are assumed to already have been trained. Second we not only need to provide a method to construct RelationDataPoint-instances but also for the fitting CoreDataPoint and all fitting CoreSlotCombinationDataPoints. This last part was somewhat difficult to do abstract, so the interface is probably not perfect. You'll get the fitting SlotSpecification, the core instance, the slot instance and the current JCas instance to construct the correct CoreSlotCombinationDataPoint. For our "IsIn" example the code could look like this:

private static class IsInDataPoint extends RelationDataPoint<Country> {

	public IsInDataPoint(Country core, ClassificationResult coreClassification,
		SlotCandidate[] slots, JCas jcas) {
		super(core, coreClassification, slots, jcas);
	}

	@Override
	public void addSpecialFeatures(FeatureMap stdFeatures, boolean addFeaturesToDictionary) {
		//we have no special features here.
	}

	private static class TrainingDataReader extends RelationTrainingDataReader<Country> {

		public TrainingDataReader(Classifier coreClassifier, Classifier[] slotClassifiers) {
			super(Country.class, coreClassifier, "ISIN");
			/*
			 * Constructing TrainingData for the RelationClassifier depends
			 * on all previous classifications. Therefore we need those as
			 * arguments here. In this case we have only one slot for the
			 * country.
			 *
			 * The Core Classifier can be directly given to the super-class
			 * using its constructor. The slots, however, have to be
			 * specified using the getSlotSpecifications().add() method.
			 */
			if (slotClassifiers.length != 1) {
				throw new UnsupportedOperationException(
					"We expected only one slot classifier!");
			}
			getSlotSpecifications().add(
				new SlotSpecification(City.class, slotClassifiers[0]));
		}

		@Override
		public RelationDataPoint<Country> createDataPoint(
			Country core, ClassificationResult coreClassification,
			SlotCandidate[] slotCandidates, JCas jcas) {
			//here we can just call the constructor
			return new IsInDataPoint(core, coreClassification, slotCandidates, jcas);
		}

		@Override
		public CoreDataPoint<Country> createCoreDataPoint(Country core, JCas jcas) {
			//here too.
			return new CountryDataPoint(core, jcas);
		}

		@Override
		public CoreSlotCombinationDataPoint<Country, ? extends Annotation>
			createCoreSlotCombinationDataPoint(
				SlotSpecification spec, Country core, Annotation slot, JCas jcas) {
					//but here we need some special handling.
				//we must choose the correct constructor ourselves.
				if (slot instanceof City) {
					return new CityDataPoint(core, (City) slot, jcas);
				} else {
					throw new UnsupportedOperationException(
						"Unexpected slot class: " + slot.getClass().getName());
				}
			}
	}
}

Training your classifiers

Training itself is fairly easy after all this previous work. You first need to instantiate a classifier. We provide an implementation based on the liblinear implementation by Benedict Waldvogel. It has some meta parameters that are described in the javadoc in more detail. You can also plug in your own classifier implementations as long as you implement the

Classifier
interface.

For our example we can read the training data and train all our classifiers with this code:

//read the raw relations from the .rel file first.
final ArrayList<RawRelation> rawRelations
	= RawRelation.readRawRelations(new File("trainData.rel"));
//then load the XCAS file.
final JCas readJCAS = JCasFactory.createJCas("src/main/resources/de/unibi/techfak/scie/descriptors/TrainExampleTypeSystem");
CasIOUtil.readXCas(readJCAS, new File("trainTestXCAS.xml"));
//then use the TrainingDataReader.
//initialize an empty training data point collection.
final ArrayList<LabeledDataPoint> countryTrainingData = new ArrayList<>();
final TrainingDataReader countryReader = new CountryDataPoint.TrainingDataReader();
countryReader.readTrainingData(jcas, rawRelations, countryTrainingData);
//now we are ready to train that classifier.
TrainingUtils.balanceDataSet(countryTrainingData);
final Classifier countryClassifier = new LibLinearClassifier(SolverType.L1R_LR);
countryClassifier.train(countryTrainingData);
SetEvaluation eval = new SetEvaluation(countryClassifier, countryTrainingData, 10);
System.out.println("");
System.out.println("--- Country Classifier Training Results ---");
System.out.println(eval.toString());
System.out.println("");
//Now we train the CityClassifier.
final ArrayList<LabeledDataPoint> cityTrainingData = new ArrayList<>();
final TrainingDataReader cityReader = new CityDataPoint.TrainingDataReader();
cityReader.readTrainingData(jcas, rawRelations, cityTrainingData);
TrainingUtils.balanceDataSet(cityTrainingData);
final Classifier cityClassifier = new LibLinearClassifier(SolverType.L1R_LR);
cityClassifier.train(cityTrainingData);
eval = new SetEvaluation(cityClassifier, cityTrainingData, 10);
System.out.println("");
System.out.println("--- City Classifier Training Results ---");
System.out.println(eval.toString());
System.out.println("");
//To train the relation classifier we need our already trained classifiers.
//other than that it is the same.
final ArrayList<LabeledDataPoint> relationTrainingData = new ArrayList<>();
final TrainingDataReader relationReader = new IsInDataPoint.TrainingDataReader(
	countryClassifier, new Classifier[]{cityClassifier});
relationReader.readTrainingData(jcas, rawRelations, relationTrainingData);
TrainingUtils.balanceDataSet(relationTrainingData);
final Classifier relationClassifier = new LibLinearClassifier(SolverType.L1R_LR);
relationClassifier.train(relationTrainingData);
eval = new SetEvaluation(relationClassifier, relationTrainingData, 10);
System.out.println("");
System.out.println("--- Relation Classifier Training Results ---");
System.out.println(eval.toString());
System.out.println("");

As you can see the process is very similar for all three types of classifier. The

TrainingUtils
provides some other methods for more sophisticated training, especially cross validation.

Creating Relation Annotations

We also provide an abstract implementation for a UIMA relation annotator based on the forementioned classifier architecture. The implementation is fairly similar to the RelationTrainingDataReader mentioned above and would look like this for our "IsIn"-example.

private static class IsInAnnotator extends RelationAnnotator<Country, IsIn> {

	public IsInAnnotator(Classifier coreClassifier,
		Classifier slotClassifier, Classifier relationClassifier) {
		super(Country.class, coreClassifier, relationClassifier);
		getSlotSpecifications().add(
			new SlotSpecification(City.class, slotClassifier));
	}

	@Override
	public RelationDataPoint<Country> createDataPoint(
		Country core, ClassificationResult coreClassification,
		SlotCandidate[] slotCandidates, JCas jcas) {
		//here we can just call the constructor
		return new IsInDataPoint(core, coreClassification, slotCandidates, jcas);
	}

	@Override
	public CoreDataPoint<Country> createCoreDataPoint(Country core, JCas jcas) {
		//here too.
		return new CountryDataPoint(core, jcas);
	}

	@Override
	public CoreSlotCombinationDataPoint<Country, ? extends Annotation>
		createCoreSlotCombinationDataPoint(
			SlotSpecification spec, Country core, Annotation slot, JCas jcas) {
				//but here we need some special handling.
			//we must choose the correct constructor ourselves.
			if (slot instanceof City) {
				return new CityDataPoint(core, (City) slot, jcas);
			} else {
				throw new UnsupportedOperationException(
					"Unexpected slot class: " + slot.getClass().getName());
			}
		}

		/*
		 * This method allows us to specify
		 * sanity checks. If they are not met we return null.
		 * In this case we want to ensure that the city slot is filled.
		 *
		 * If it is filled we create an isIn annotation from it.
		 * We do not need to set begin and end. The super-class will
		 * take care of that. But we need to fill the slots.
		 */
		@Override
		public IsIn createRelationObject(
			Country core, ClassificationResult coreClassification,
			SlotCandidate[] slots, JCas jcas) {
			if (slots[0] == null || slots[0].getAnnotation() == null) {
				return null;
			}
			final City city = (City) slots[0].getAnnotation();
			final IsIn relationInstance = new IsIn(jcas);
			relationInstance.setSubject(city);
			relationInstance.setObject(core);
			return relationInstance;
		}
}

The main difference is the

createRelationObject
method that allows to implement sanity checks. If they are not met we can return null even though all classifiers have accepted this relation instance. In this case we use this mechanism to enforce that both slots are filled.

The code of the whole example can be found at:

src/test/java/de/unibi/techfak/scie/classifiers/RelationExtractionTest.java

Dependencies

This software depends on the uimaj-core and the uimafit-core of Apache UIMA and on the liblinear implementation of Benedict Waldvogel. Former is used according to the terms of the Apache License, Version 2 while latter has a custom license. In accordance to this license we hereby provide the copyright notice:

Copyright (c) 2007-2013 The LIBLINEAR Project.
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:

1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.

2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.

3. Neither name of copyright holders nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.


THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

GNU AFFERO GENERAL PUBLIC LICENSE

Version 3, 19 November 2007

Copyright © 2007 Free Software Foundation, Inc. <http://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.

Preamble

The GNU Affero General Public License is a free, copyleft license for software and other kinds of works, specifically designed to ensure cooperation with the community in the case of network server software.

The licenses for most software and other practical works are designed to take away your freedom to share and change the works. By contrast, our General Public Licenses are intended to guarantee your freedom to share and change all versions of a program--to make sure it remains free software for all its users.

When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for them if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs, and that you know you can do these things.

Developers that use our General Public Licenses protect your rights with two steps: (1) assert copyright on the software, and (2) offer you this License which gives you legal permission to copy, distribute and/or modify the software.

A secondary benefit of defending all users' freedom is that improvements made in alternate versions of the program, if they receive widespread use, become available for other developers to incorporate. Many developers of free software are heartened and encouraged by the resulting cooperation. However, in the case of software used on network servers, this result may fail to come about. The GNU General Public License permits making a modified version and letting the public access it on a server without ever releasing its source code to the public.

The GNU Affero General Public License is designed specifically to ensure that, in such cases, the modified source code becomes available to the community. It requires the operator of a network server to provide the source code of the modified version running there to the users of that server. Therefore, public use of a modified version, on a publicly accessible server, gives the public access to the source code of the modified version.

An older license, called the Affero General Public License and published by Affero, was designed to accomplish similar goals. This is a different license, not a version of the Affero GPL, but Affero has released a new version of the Affero GPL which permits relicensing under this license.

The precise terms and conditions for copying, distribution and modification follow.

TERMS AND CONDITIONS

0. Definitions.

"This License" refers to version 3 of the GNU Affero General Public License.

"Copyright" also means copyright-like laws that apply to other kinds of works, such as semiconductor masks.

"The Program" refers to any copyrightable work licensed under this License. Each licensee is addressed as "you". "Licensees" and "recipients" may be individuals or organizations.

To "modify" a work means to copy from or adapt all or part of the work in a fashion requiring copyright permission, other than the making of an exact copy. The resulting work is called a "modified version" of the earlier work or a work "based on" the earlier work.

A "covered work" means either the unmodified Program or a work based on the Program.

To "propagate" a work means to do anything with it that, without permission, would make you directly or secondarily liable for infringement under applicable copyright law, except executing it on a computer or modifying a private copy. Propagation includes copying, distribution (with or without modification), making available to the public, and in some countries other activities as well.

To "convey" a work means any kind of propagation that enables other parties to make or receive copies. Mere interaction with a user through a computer network, with no transfer of a copy, is not conveying.

An interactive user interface displays "Appropriate Legal Notices" to the extent that it includes a convenient and prominently visible feature that (1) displays an appropriate copyright notice, and (2) tells the user that there is no warranty for the work (except to the extent that warranties are provided), that licensees may convey the work under this License, and how to view a copy of this License. If the interface presents a list of user commands or options, such as a menu, a prominent item in the list meets this criterion.

1. Source Code.

The "source code" for a work means the preferred form of the work for making modifications to it. "Object code" means any non-source form of a work.

A "Standard Interface" means an interface that either is an official standard defined by a recognized standards body, or, in the case of interfaces specified for a particular programming language, one that is widely used among developers working in that language.

The "System Libraries" of an executable work include anything, other than the work as a whole, that (a) is included in the normal form of packaging a Major Component, but which is not part of that Major Component, and (b) serves only to enable use of the work with that Major Component, or to implement a Standard Interface for which an implementation is available to the public in source code form. A "Major Component", in this context, means a major essential component (kernel, window system, and so on) of the specific operating system (if any) on which the executable work runs, or a compiler used to produce the work, or an object code interpreter used to run it.

The "Corresponding Source" for a work in object code form means all the source code needed to generate, install, and (for an executable work) run the object code and to modify the work, including scripts to control those activities. However, it does not include the work's System Libraries, or general-purpose tools or generally available free programs which are used unmodified in performing those activities but which are not part of the work. For example, Corresponding Source includes interface definition files associated with source files for the work, and the source code for shared libraries and dynamically linked subprograms that the work is specifically designed to require, such as by intimate data communication or control flow between those subprograms and other parts of the work.

The Corresponding Source need not include anything that users can regenerate automatically from other parts of the Corresponding Source.

The Corresponding Source for a work in source code form is that same work.

2. Basic Permissions.

All rights granted under this License are granted for the term of copyright on the Program, and are irrevocable provided the stated conditions are met. This License explicitly affirms your unlimited permission to run the unmodified Program. The output from running a covered work is covered by this License only if the output, given its content, constitutes a covered work. This License acknowledges your rights of fair use or other equivalent, as provided by copyright law.

You may make, run and propagate covered works that you do not convey, without conditions so long as your license otherwise remains in force. You may convey covered works to others for the sole purpose of having them make modifications exclusively for you, or provide you with facilities for running those works, provided that you comply with the terms of this License in conveying all material for which you do not control copyright. Those thus making or running the covered works for you must do so exclusively on your behalf, under your direction and control, on terms that prohibit them from making any copies of your copyrighted material outside their relationship with you.

Conveying under any other circumstances is permitted solely under the conditions stated below. Sublicensing is not allowed; section 10 makes it unnecessary.

3. Protecting Users' Legal Rights From Anti-Circumvention Law.

No covered work shall be deemed part of an effective technological measure under any applicable law fulfilling obligations under article 11 of the WIPO copyright treaty adopted on 20 December 1996, or similar laws prohibiting or restricting circumvention of such measures.

When you convey a covered work, you waive any legal power to forbid circumvention of technological measures to the extent such circumvention is effected by exercising rights under this License with respect to the covered work, and you disclaim any intention to limit operation or modification of the work as a means of enforcing, against the work's users, your or third parties' legal rights to forbid circumvention of technological measures.

4. Conveying Verbatim Copies.

You may convey verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice; keep intact all notices stating that this License and any non-permissive terms added in accord with section 7 apply to the code; keep intact all notices of the absence of any warranty; and give all recipients a copy of this License along with the Program.

You may charge any price or no price for each copy that you convey, and you may offer support or warranty protection for a fee.

5. Conveying Modified Source Versions.

You may convey a work based on the Program, or the modifications to produce it from the Program, in the form of source code under the terms of section 4, provided that you also meet all of these conditions:

A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an "aggregate" if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate.

6. Conveying Non-Source Forms.

You may convey a covered work in object code form under the terms of sections 4 and 5, provided that you also convey the machine-readable Corresponding Source under the terms of this License, in one of these ways:

A separable portion of the object code, whose source code is excluded from the Corresponding Source as a System Library, need not be included in conveying the object code work.

A "User Product" is either (1) a "consumer product", which means any tangible personal property which is normally used for personal, family, or household purposes, or (2) anything designed or sold for incorporation into a dwelling. In determining whether a product is a consumer product, doubtful cases shall be resolved in favor of coverage. For a particular product received by a particular user, "normally used" refers to a typical or common use of that class of product, regardless of the status of the particular user or of the way in which the particular user actually uses, or expects or is expected to use, the product. A product is a consumer product regardless of whether the product has substantial commercial, industrial or non-consumer uses, unless such uses represent the only significant mode of use of the product.

"Installation Information" for a User Product means any methods, procedures, authorization keys, or other information required to install and execute modified versions of a covered work in that User Product from a modified version of its Corresponding Source. The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made.

If you convey an object code work under this section in, or with, or specifically for use in, a User Product, and the conveying occurs as part of a transaction in which the right of possession and use of the User Product is transferred to the recipient in perpetuity or for a fixed term (regardless of how the transaction is characterized), the Corresponding Source conveyed under this section must be accompanied by the Installation Information. But this requirement does not apply if neither you nor any third party retains the ability to install modified object code on the User Product (for example, the work has been installed in ROM).

The requirement to provide Installation Information does not include a requirement to continue to provide support service, warranty, or updates for a work that has been modified or installed by the recipient, or for the User Product in which it has been modified or installed. Access to a network may be denied when the modification itself materially and adversely affects the operation of the network or violates the rules and protocols for communication across the network.

Corresponding Source conveyed, and Installation Information provided, in accord with this section must be in a format that is publicly documented (and with an implementation available to the public in source code form), and must require no special password or key for unpacking, reading or copying.

7. Additional Terms.

"Additional permissions" are terms that supplement the terms of this License by making exceptions from one or more of its conditions. Additional permissions that are applicable to the entire Program shall be treated as though they were included in this License, to the extent that they are valid under applicable law. If additional permissions apply only to part of the Program, that part may be used separately under those permissions, but the entire Program remains governed by this License without regard to the additional permissions.

When you convey a copy of a covered work, you may at your option remove any additional permissions from that copy, or from any part of it. (Additional permissions may be written to require their own removal in certain cases when you modify the work.) You may place additional permissions on material, added by you to a covered work, for which you have or can give appropriate copyright permission.

Notwithstanding any other provision of this License, for material you add to a covered work, you may (if authorized by the copyright holders of that material) supplement the terms of this License with terms:

All other non-permissive additional terms are considered "further restrictions" within the meaning of section 10. If the Program as you received it, or any part of it, contains a notice stating that it is governed by this License along with a term that is a further restriction, you may remove that term. If a license document contains a further restriction but permits relicensing or conveying under this License, you may add to a covered work material governed by the terms of that license document, provided that the further restriction does not survive such relicensing or conveying.

If you add terms to a covered work in accord with this section, you must place, in the relevant source files, a statement of the additional terms that apply to those files, or a notice indicating where to find the applicable terms.

Additional terms, permissive or non-permissive, may be stated in the form of a separately written license, or stated as exceptions; the above requirements apply either way.

8. Termination.

You may not propagate or modify a covered work except as expressly provided under this License. Any attempt otherwise to propagate or modify it is void, and will automatically terminate your rights under this License (including any patent licenses granted under the third paragraph of section 11).

However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation.

Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice.

Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, you do not qualify to receive new licenses for the same material under section 10.

9. Acceptance Not Required for Having Copies.

You are not required to accept this License in order to receive or run a copy of the Program. Ancillary propagation of a covered work occurring solely as a consequence of using peer-to-peer transmission to receive a copy likewise does not require acceptance. However, nothing other than this License grants you permission to propagate or modify any covered work. These actions infringe copyright if you do not accept this License. Therefore, by modifying or propagating a covered work, you indicate your acceptance of this License to do so.

10. Automatic Licensing of Downstream Recipients.

Each time you convey a covered work, the recipient automatically receives a license from the original licensors, to run, modify and propagate that work, subject to this License. You are not responsible for enforcing compliance by third parties with this License.

An "entity transaction" is a transaction transferring control of an organization, or substantially all assets of one, or subdividing an organization, or merging organizations. If propagation of a covered work results from an entity transaction, each party to that transaction who receives a copy of the work also receives whatever licenses to the work the party's predecessor in interest had or could give under the previous paragraph, plus a right to possession of the Corresponding Source of the work from the predecessor in interest, if the predecessor has it or can get it with reasonable efforts.

You may not impose any further restrictions on the exercise of the rights granted or affirmed under this License. For example, you may not impose a license fee, royalty, or other charge for exercise of rights granted under this License, and you may not initiate litigation (including a cross-claim or counterclaim in a lawsuit) alleging that any patent claim is infringed by making, using, selling, offering for sale, or importing the Program or any portion of it.

11. Patents.

A "contributor" is a copyright holder who authorizes use under this License of the Program or a work on which the Program is based. The work thus licensed is called the contributor's "contributor version".

A contributor's "essential patent claims" are all patent claims owned or controlled by the contributor, whether already acquired or hereafter acquired, that would be infringed by some manner, permitted by this License, of making, using, or selling its contributor version, but do not include claims that would be infringed only as a consequence of further modification of the contributor version. For purposes of this definition, "control" includes the right to grant patent sublicenses in a manner consistent with the requirements of this License.

Each contributor grants you a non-exclusive, worldwide, royalty-free patent license under the contributor's essential patent claims, to make, use, sell, offer for sale, import and otherwise run, modify and propagate the contents of its contributor version.

In the following three paragraphs, a "patent license" is any express agreement or commitment, however denominated, not to enforce a patent (such as an express permission to practice a patent or covenant not to sue for patent infringement). To "grant" such a patent license to a party means to make such an agreement or commitment not to enforce a patent against the party.

If you convey a covered work, knowingly relying on a patent license, and the Corresponding Source of the work is not available for anyone to copy, free of charge and under the terms of this License, through a publicly available network server or other readily accessible means, then you must either (1) cause the Corresponding Source to be so available, or (2) arrange to deprive yourself of the benefit of the patent license for this particular work, or (3) arrange, in a manner consistent with the requirements of this License, to extend the patent license to downstream recipients. "Knowingly relying" means you have actual knowledge that, but for the patent license, your conveying the covered work in a country, or your recipient's use of the covered work in a country, would infringe one or more identifiable patents in that country that you have reason to believe are valid.

If, pursuant to or in connection with a single transaction or arrangement, you convey, or propagate by procuring conveyance of, a covered work, and grant a patent license to some of the parties receiving the covered work authorizing them to use, propagate, modify or convey a specific copy of the covered work, then the patent license you grant is automatically extended to all recipients of the covered work and works based on it.

A patent license is "discriminatory" if it does not include within the scope of its coverage, prohibits the exercise of, or is conditioned on the non-exercise of one or more of the rights that are specifically granted under this License. You may not convey a covered work if you are a party to an arrangement with a third party that is in the business of distributing software, under which you make payment to the third party based on the extent of your activity of conveying the work, and under which the third party grants, to any of the parties who would receive the covered work from you, a discriminatory patent license (a) in connection with copies of the covered work conveyed by you (or copies made from those copies), or (b) primarily for and in connection with specific products or compilations that contain the covered work, unless you entered into that arrangement, or that patent license was granted, prior to 28 March 2007.

Nothing in this License shall be construed as excluding or limiting any implied license or other defenses to infringement that may otherwise be available to you under applicable patent law.

12. No Surrender of Others' Freedom.

If conditions are imposed on you (whether by court order, agreement or otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of this License. If you cannot convey a covered work so as to satisfy simultaneously your obligations under this License and any other pertinent obligations, then as a consequence you may not convey it at all. For example, if you agree to terms that obligate you to collect a royalty for further conveying from those to whom you convey the Program, the only way you could satisfy both those terms and this License would be to refrain entirely from conveying the Program.

13. Remote Network Interaction; Use with the GNU General Public License.

Notwithstanding any other provision of this License, if you modify the Program, your modified version must prominently offer all users interacting with it remotely through a computer network (if your version supports such interaction) an opportunity to receive the Corresponding Source of your version by providing access to the Corresponding Source from a network server at no charge, through some standard or customary means of facilitating copying of software. This Corresponding Source shall include the Corresponding Source for any work covered by version 3 of the GNU General Public License that is incorporated pursuant to the following paragraph.

Notwithstanding any other provision of this License, you have permission to link or combine any covered work with a work licensed under version 3 of the GNU General Public License into a single combined work, and to convey the resulting work. The terms of this License will continue to apply to the part which is the covered work, but the work with which it is combined will remain governed by version 3 of the GNU General Public License.

14. Revised Versions of this License.

The Free Software Foundation may publish revised and/or new versions of the GNU Affero General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns.

Each version is given a distinguishing version number. If the Program specifies that a certain numbered version of the GNU Affero General Public License "or any later version" applies to it, you have the option of following the terms and conditions either of that numbered version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of the GNU Affero General Public License, you may choose any version ever published by the Free Software Foundation.

If the Program specifies that a proxy can decide which future versions of the GNU Affero General Public License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Program.

Later license versions may give you additional or different permissions. However, no additional obligations are imposed on any author or copyright holder as a result of your choosing to follow a later version.

15. Disclaimer of Warranty.

THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.

16. Limitation of Liability.

IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

17. Interpretation of Sections 15 and 16.

If the disclaimer of warranty and limitation of liability provided above cannot be given local legal effect according to their terms, reviewing courts shall apply local law that most closely approximates an absolute waiver of all civil liability in connection with the Program, unless a warranty or assumption of liability accompanies a copy of the Program in return for a fee.

END OF TERMS AND CONDITIONS

How to Apply These Terms to Your New Programs

If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms.

To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively state the exclusion of warranty; and each file should have at least the "copyright" line and a pointer to where the full notice is found.

    <one line to give the program's name and a brief idea of what it does.>
    Copyright (C) <year>  <name of author>

    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU Affero General Public License as
    published by the Free Software Foundation, either version 3 of the
    License, or (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU Affero General Public License for more details.

    You should have received a copy of the GNU Affero General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.

Also add information on how to contact you by electronic and paper mail.

If your software can interact with users remotely through a computer network, you should also make sure that it provides a way for users to get its source. For example, if your program is a web application, its interface could display a "Source" link that leads users to an archive of the code. There are many ways you could offer source, and different solutions will be better for different programs; see section 13 for the specific requirements.

You should also get your employer (if you work as a programmer) or school, if any, to sign a "copyright disclaimer" for the program, if necessary. For more information on this, and how to apply and follow the GNU AGPL, see <http://www.gnu.org/licenses/>.