HebMorph enables high‑quality Hebrew search with Apache Lucene by providing tokenization, normalization, and morphology support tailored for Hebrew. The project aims for strong recall and precision, and is released under the GNU Affero General Public License v3.

More details: http://code972.com/HebMorph

Note: The .NET codebase and legacy components were removed. The maintained codebase is the Java Lucene module published as hebmorph-lucene.

• HebMorph releases track Apache Lucene’s versions. Artifact versions generally match the Lucene version they target.
• For Elasticsearch, use the dedicated plugin: https://github.com/synhershko/elasticsearch-analysis-hebrew

This project is not published to Maven Central yet. Use one of the following options:

• Build from source and depend on modules directly.

• Publish to your local Maven cache and depend on it from your app:

  1. Publish locally

  ./gradlew publishToMavenLocal

  2. Add mavenLocal() to your project repositories and depend on the artifact:

  • Gradle (Kotlin DSL)

    repositories { mavenLocal(); mavenCentral() } dependencies { implementation("com.code972.hebmorph:hebmorph-lucene:10.3.1") }

  • Maven

    com.code972.hebmorph hebmorph-lucene 10.3.1

  The artifact version is aligned with Lucene’s version (see gradle/libs.versions.toml).

• Optional: configure a remote repository (e.g., GitHub Packages) in publishing.repositories if you want to distribute a private or public build.

• Java 21+
• HSpell data files available locally (the hspell-data-files directory).

At runtime or during tests, HebMorph tries to resolve the path to hspell-data-files dynamically. The resolution order is:

1. Java system property: -Dhebmorph.hspell.path=/path/to/hspell-data-files
2. Environment variable: HEBMORPH_HSPELL_PATH=/path/to/hspell-data-files
3. Common relative paths near the project (e.g. ../hspell-data-files)

If none is set and the files are not found, tests will fail with a clear error. Place the hspell-data-files directory next to the repository root or set one of the overrides above.

This project uses Gradle with Kotlin DSL and a centralized version catalog (gradle/libs.versions.toml).

• Build: ./gradlew build
• Run tests (ensure HSpell data path is configured as above): ./gradlew test

The primary analyzers are under org.apache.lucene.analysis.hebrew:

• HebrewIndexingAnalyzer for indexing
• HebrewQueryAnalyzer and HebrewQueryLightAnalyzer for queries
• HebrewExactAnalyzer for exact match use‑cases

See the test sources for usage examples.

• Build migrated to Gradle Kotlin DSL; configuration streamlined.
• Centralized dependency management via gradle/libs.versions.toml.
• Modernized Java target (Java 21) and updated Lucene dependency.
• Dynamic resolution of hspell-data-files path via system property or environment variable.
• Removed legacy .NET code and outdated components.

HebMorph is copyright (C) 2010–2015, Itamar Syn‑Hershko. HebMorph relies on Hspell, copyright (C) 2000–2013, Nadav Har'El and Dan Kenigsberg (http://hspell.ivrix.org.il/).

It is released under the GNU Affero General Public License v3. See the LICENSE file. Note that not only the programs in the distribution, but also the dictionary files and the generated word lists, are licensed under the AGPL. There is no warranty of any kind for the contents of this distribution.