Welcome to the Apache UIMA Java SDK

Apache UIMA helps you managing unstructured data (such as texts) that is enriched useful information. For example, if you want to identify a mention of an entity in a text or possible link that entity to a reference dataset, then Apache UIMA provides:

• a convenient data structure --the Common Analysis Structure (CAS)-- to represent that data
• a type system concept service as a schema for the enriched data that is stored in the CAS
• a component model consisting of reader, analysis engines (processors) and consumers (writers) to process that data
• a model for aggregating multiple analysis engines into pipelines and executing them (optionally parallelized)
• various options for (de)serializing the CAS from/to different formats
• any many additional features!

Note the Apache UIMA Java SDK only provides a framework for building analytics but it does not provide any analytics. However, there are various third-parties that build on Apache UIMA and that provide collections of analysis components or ready-made solutions.

System requirements

Apache UIMA v3.5.0 and later requires Java version 17 or later.

Running the Eclipse plugin tooling for UIMA requires you start Eclipse 4.25 (2022-09) or later using a Java 17 or later.

Running the migration tool on class files requires running with a Java JDK, not a Java JRE.

The supported platforms are: Windows, Linux, and macOS. Other Java platform implementations should work but have not been significantly tested.

Many of the scripts in the /bin directory invoke Java. They use the value of the environment variable, JAVA_HOME, to locate the Java to use; if it is not set, they invoke java expecting to find an appropriate Java in your PATH variable.

Using Apache UIMA Java SDK

You can add the Apache UIMA Java SDK to your project easily in most build tools by importing it from Maven Central. For example if you use Maven, you can add the following dependency to your project:

<dependency>
  <groupId>org.apache.uima</groupId>
  <artifactId>uimaj-core</artifactId>
  <version>3.5.0</version>
</dependency>

Next, we give a few brief examples of how to use the Apache UIMA Java SDK and the Apache uimaFIT library. Apache uimaFIT is a separate dependency that you can add:

<dependency>
  <groupId>org.apache.uima</groupId>
  <artifactId>uimafit-core</artifactId>
  <version>3.5.0</version>
</dependency>

Creating a type system

The type system defines the type of information that we want to attach to the unstructured information (here a text document). In our example, we want to identify mentions of entities, so we define a type my.Entity with a feature category which can be used to store the category the entity belongs to.

To illustrate the information UIMA internally maintains about the annotation schema, we write the generated schema as XML to screen.

String TYPE_NAME_ENTITY = "my.Entity";
String TYPE_NAME_TOKEN = "my.Token";
String FEAT_NAME_CATEGORY = "category";

var tsd = UIMAFramework.getResourceSpecifierFactory().createTypeSystemDescription();
tsd.addType(TYPE_NAME_TOKEN, "", CAS.TYPE_NAME_ANNOTATION);
var entityTypeDesc = tsd.addType(TYPE_NAME_ENTITY, "", CAS.TYPE_NAME_ANNOTATION);
entityTypeDesc.addFeature(FEAT_NAME_CATEGORY, "", CAS.TYPE_NAME_STRING);

tsd.toXML(System.out);

Creating a Common Analaysis Structure object

Now we create a Common Analysis Structure (CAS) object into which we store the text that we want to analyse.

Again, to illustrate the information that UIMA internally stores in the CAS object, we write an XML representation of the object to screen.

var cas = CasFactory.createCas(tsd);
cas.setDocumentText("Welcome to Apache UIMA.");
cas.setDocumentLanguage("en");

CasIOUtils.save(cas, System.out, SerialFormat.XMI_PRETTY);

Adding and retrieving annotations

Now, we create an annotation of the type my.Entity to identify the mention of Apache UIMA in the example text.

Finally, we iterate over all annotations in the CAS and print them to screen. This includes the default DocumentAnnotation that is always created by UIMA as well as the my.Entity annotation that we created ourselves.

var entityType = cas.getTypeSystem().getType(TYPE_NAME_ENTITY);
var entity = cas.createAnnotation(entityType, 11, 22);
cas.addFsToIndexes(entity);

for (var anno : cas.<Annotation>select(entityType)) {
   System.out.printf("%s: [%s]%n", anno.getType().getName(), anno.getCoveredText());
}

Working with analysis components

In order to organize different types of analysis into steps, we usually package them into individual analysis engines. We illustrate now how such components can be built and how they can be put executed as an analysis pipeline.

class TokenAnnotator extends CasAnnotator_ImplBase {
  public void process(CAS cas) throws AnalysisEngineProcessException {
    var tokenType = cas.getTypeSystem().getType(TYPE_NAME_TOKEN);
    var bi = BreakIterator.getWordInstance();
    bi.setText(cas.getDocumentText());
    int begin = bi.first();
    int end;
    for (end = bi.next(); end != BreakIterator.DONE; end = bi.next()) {
      var token = cas.createAnnotation(tokenType, begin, end);
      cas.addFsToIndexes(token);
      begin = end;
    }
  }
}

class EntityAnnotator extends CasAnnotator_ImplBase {
  public void process(CAS cas) throws AnalysisEngineProcessException {
    var tokenType = cas.getTypeSystem().getType(TYPE_NAME_TOKEN);
    var entityType = cas.getTypeSystem().getType(TYPE_NAME_ENTITY);
    for (var token : cas.<Annotation>select(tokenType)) {
      if (Character.isUpperCase(token.getCoveredText().charAt(0))) {
        var entity = cas.createAnnotation(entityType, token.getBegin(), token.getEnd());
        cas.addFsToIndexes(entity);
      }
    }
  }
}

cas = CasFactory.createCas(tsd);
cas.setDocumentText("John likes Apache UIMA.");
cas.setDocumentLanguage("en");

var pipeline = AnalysisEngineFactory.createEngineDescription(
  AnalysisEngineFactory.createEngineDescription(TokenAnnotator.class),
  AnalysisEngineFactory.createEngineDescription(EntityAnnotator.class));

SimplePipeline.runPipeline(cas, pipeline);

for (var anno : cas.<Annotation>select(entityType)) {
   System.out.printf("%s: [%s]%n", anno.getType().getName(), anno.getCoveredText());
}

Using uimaFIT

Configuring UIMA components is generally achieved by creating XML descriptor files which tell the framework at runtime how components should be instantiated and deployed. These XML descriptor files are very tightly coupled with the Java implementation of the components they describe. We have found that it is very difficult to keep the two consistent with each other especially when code refactoring is very frequent. uimaFIT provides Java annotations for describing UIMA components which can be used to directly describe the UIMA components in the code. This greatly simplifies refactoring a component definition (e.g. changing a configuration parameter name). It also makes it possible to generate XML descriptor files as part of the build cycle rather than being performed manually in parallel with code creation. uimaFIT also makes it easy to instantiate UIMA components without using XML descriptor files at all by providing a number of convenience factory methods which allow programmatic/dynamic instantiation of UIMA components. This makes uimaFIT an ideal library for testing UIMA components because the component can be easily instantiated and invoked without requiring a descriptor file to be created first. uimaFIT is also helpful in research environments in which programmatic/dynamic instantiation of a pipeline can simplify experimentation. For example, when performing 10-fold cross-validation across a number of experimental conditions it can be quite laborious to create a different set of descriptor files for each run or even a script that generates such descriptor files. uimaFIT is type system agnostic and does not depend on (or provide) a specific type system.

uimaFIT is a library that provides factories, injection, and testing utilities for UIMA. The following list highlights some of the features uimaFIT provides:

• Factories: simplify instantiating UIMA components programmatically without descriptor files. For example, to instantiate an AnalysisEngine a call like this could be made:

  AnalysisEngineFactory.createEngine(MyAEImpl.class, myTypeSystem,
    paramName1, paramValue2, 
    paramName2, paramValue2, 
    ...)
  

• Injection: handles the binding of configuration parameter values to the corresponding member variables in the analysis engines and handles the binding of external resources. For example, to bind a configuration parameter just annotate a member variable with @ConfigurationParameter. External resources can likewise by injected via the @ExternalResource annotation. Then add one line of code to your initialize method:

  ConfigurationParameterInitializer.initialize(this, uimaContext).
  

  This is handled automatically if you extend the uimaFIT JCasAnnotator_ImplBase class.

• Testing: uimaFIT simplifies testing in a number of ways described in the documentation. By making it easy to instantiate your components without descriptor files a large amount of difficult-to-maintain and unnecessary XML can be eliminated from your test code. This makes tests easier to write and maintain. Also, running components as a pipeline can be accomplished with a method call like this:

  SimplePipeline.runPipeline(reader, ae1, ..., aeN, consumer1, ... consumerN)
  

uimaFIT is a part of the Apache UIMA(TM) project. uimaFIT can only be used in conjunction with a compatible version of the Java version of the Apache UIMA SDK. For your convenience, the binary distribution package of uimaFIT includes all libraries necessary to use uimaFIT. In particular for novice users, it is strongly advised to obtain a copy of the full UIMA SDK separately.

uimaFIT is available via Maven Central. If you use Maven for your build environment, then you can add uimaFIT as a dependency to your pom.xml file with the following:

<dependencies>
  <dependency>
    <groupId>org.apache.uima</groupId>
    <artifactId>uimafit-core</artifactId>
    <version>3.5.0</version>
  </dependency>
</dependencies>

Modules

• uimafit-core - the main uimaFIT module
• uimafit-cpe - support for the Collection Processing Engine (multi-threaded pipelines)
• uimafit-maven - a Maven plugin to automatically enhance UIMA components with uimaFIT metadata and to generate XML descriptors for uimaFIT-enabled components.
• uimafit-junit - convenience code facilitating the implementation of UIMA/ uimaFIT tests in JUnit tests
• uimafit-assertj - adds assertions for UIMA/uimaFIT types via the AssertJ framework
• uimafit-spring - an experimental module serving as a proof-of-concept for the integration of UIMA with the Spring Framework. It is currently not considered finished and uses invasive reflection in order to patch the UIMA framework such that it passes all components created by UIMA through Spring to provide for the wiring of Spring context dependencies. This module is made available for the adventurous but currently not considered stable, finished, or even a proper part of the package. E.g. it is not included in the binary distribution package.

Building

To build Apache UIMA, you need at least a Java 17 JDK and a recent Maven 3 version.

After extracting the source distribution ZIP or cloning the repository, change into the created directory and run the following command:

mvn clean install

For more details, please see http://uima.apache.org/building-uima.html

Running examples from the source/binary distribution

You can download the source and binary distributions from the Apache UIMA website.

Environment Variables

After you have unpacked the Apache UIMA distribution from the package of your choice (e.g. .zip or .gz), perform the steps below to set up UIMA so that it will function properly.

• Set JAVA_HOME to the directory of your JRE installation you would like to use for UIMA.

• Set UIMA_HOME to the apache-uima directory of your unpacked Apache UIMA distribution

• Append UIMA_HOME/bin to your PATH

• Please run the script UIMA_HOME/bin/adjustExamplePaths.bat (or .sh), to update paths in the examples based on the actual UIMA_HOME directory path. This script runs a Java program; you must either have java in your PATH or set the environment variable JAVA_HOME to a suitable JRE.

  Note: The Mac OS X operating system procedures for setting up global environment variables are described here: see http://developer.apple.com/qa/qa2001/qa1067.html.

Verifying Your Installation

To test the installation, run the documentAnalyzer.bat (or .sh) file located in the bin subdirectory. This should pop up a Document Analyzer window. Set the values displayed in this GUI to as follows:

• Input Directory: UIMA_HOME/examples/data
• Output Directory: UIMA_HOME/examples/data/processed
• Location of Analysis Engine XML Descriptor: UIMA_HOME/examples/descriptors/analysis_engine/PersonTitleAnnotator.xml

Replace UIMA_HOME above with the path of your Apache UIMA installation.

Next, click the Run button, which should, after a brief pause, pop up an Analyzed Results window. Double-click on one of the documents to display the analysis results for that document.

UIMA component providers

Here is list of several well-known projects that provide their analysis tools as UIMA components or that wrap third-party analysis tools as UIMA components:

• Apache cTAKES - Natural language processing system for extraction of information from electronic medical record clinical free-text.
• Apache OpenNLP - Wraps OpenNLP for UIMA. Adaptable to different type systems.
• Apache Ruta - Generic rule-based text analytics. Works with any type system.
• ClearTK - Wraps several third-party tools (OpenNLP, CoreNLP, etc.) and offers a flexible framework for training own machine learning models. Uses CleartK type system.
• DKPro Core - Wraps many third-party tools (OpenNLP, CoreNLP, etc.) and supporting a wide range of data formats. Uses DKPro Core type system.
• JULIE Lab Component Repository (JCoRe) Wraps several third-party tools (OpenNLP, CoreNLP, etc.) and supporting a wide range of data formats, in particular from the biomed domain. Uses JCore type system.

This is not an exhaustive list. If you feel any particular project should be listed here, please let us know. You could find additional ones e.g. by:

• following the GitHub dependency graph
• searching Google Scholar for UIMA

Interoperability

The Apache UIMA Java SDK can be used with any programming language based on the Java Virtual Machine including Java, Groovy, Scala, and many other languages.

Interoperability with Python can for example be achieved via the third-party DKPro Cassis library which can be used to read, manipulate and write CAS data in the XMI format.

Further reading

The Apache UIMA Java SDK is a Java-based implementation of the UIMA specification.

Support

Please direct questions to [email protected].

Reference

If you use uimaFIT to support academic research, then please consider citing the following paper as appropriate:

@InProceedings{ogren-bethard:2009:SETQA-NLP,
  author    = {Ogren, Philip  and  Bethard, Steven},
  title     = {Building Test Suites for {UIMA} Components},
  booktitle = {Proceedings of the Workshop on Software Engineering, Testing, and Quality Assurance for Natural Language Processing (SETQA-NLP 2009)},
  month     = {June},
  year      = {2009},
  address   = {Boulder, Colorado},
  publisher = {Association for Computational Linguistics},
  pages     = {1--4},
  url       = {http://www.aclweb.org/anthology/W/W09/W09-1501}
}

History

• Early 2000s: UIMA was originally developed by IBM as part of research into analyzing unstructured information (like text, audio, and video). It was designed to process large volumes of unstructured data in a scalable way, targeting natural language processing (NLP) applications.

• 2004: UIMA was open-sourced allowing for broader use and contributions from outside IBM.

• 2006: The UIMA project was accepted into the Apache Incubator, starting the formal process of becoming an Apache project.

• 2008: UIMA graduated from the Apache Incubator and became a top-level Apache project, signifying its maturity and active development.

• 2009: Apache UIMA-AS (Asynchronous Scaleout) was introduced, enabling distributed and asynchronous processing of UIMA pipelines.

• 2012: uimaFIT was contributed to the Apache UIMA project. Apache uimaFIT was formerly known as uimaFIT, which in turn was formerly known as UUTUC. Prior to its contribution, is was collaborative effort between the Center for Computational Pharmacology at the University of Colorado Denver, the Center for Computational Language and Education Research at the University of Colorado at Boulder, and the Ubiquitous Knowledge Processing (UKP) Lab at the Technische Universität Darmstadt.

• 2013: UIMA DUCC (Distributed UIMA Cluster Computing) was introduced as a sub-project of Apache UIMA.

• 2016: Apache UIMA Ruta (Rule-based Text Annotation) was introduced as an extension, providing a scripting language for rule-based text processing.

• 2023: UIMA DUCC and UIMA-AS were retired.

• 2024: uimaFIT has been merged into the UIMA Java SDK