/annotace

Annotace is a text annotation service for Czech and English. It is used for support annotations in TermIt

Primary LanguageHTMLGNU General Public License v3.0GPL-3.0

Annotace

Annotace is a text analysis service used e.g. by TermIt and its web annotation plugin.

How to run it?

  • Install Java 11
  • Run ./gradlew bootRun (on Linux/WSL) or gradlew.bat bootRun on Windows

Lemmatizers

Annotace supports two lemmatizer implementations:

  • Spark-based lemmatizer is more suitable for annotation of English texts. This is the default lemmatizer
  • MorphoDiTa-based lemmatizer is more suitable for annotation of Czech or Slovak texts. It comes in two variants:
    • JNI-based - runs locally using the MorphoDiTa library itself
    • Service-based - invokes a remote annotation service (needs to be configured)

Setup

Spark-based Annotace setup does not require any additional configuration or files. Either run it directly ./gradlew bootRun or use Docker. There is an image published at GitHub package registry.

Running Annotace with MorphoDiTa is a bit more complicated.

Annotace with MorphoDiTa Locally

  1. Download the MorphoDiTa ZIP archive and extract it.
  2. Find a file with JNI bindings corresponding to your platform in the extracted directory. For 64-bit Linux the file is morphodita-1.9.2-bin/bin-linux64/java/libmorphodita_java.so.
  3. Set path to the directory containing this file as java.library.path environment variable name.
  4. Provide mapping of taggers (language models) to Annotace. Either by editing application.yml before build or by passing them as environment variables.
  5. Run Annotace with the MorphoDiTa lemmatizer by setting ANNOTACE_LEMMATIZER to morphodita-jni.

A complete command line example would be: ANNOTACE_LEMMATIZER=morphodita-jni ANNOTACE_MORPHODITA_TAGGERS_CS=/opt/annotace/lib/czech-morfflex2.0-pdtc1.0-220710.tagger LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/annotace/lib/morphodita-1.9.2-bin/bin-linux64/java ./gradlew bootRun

Annotace with MorphoDiTa in Docker

  1. Download the MorphoDiTa ZIP archive.
  2. Set MORPHODITA_ZIP in docker-compose-morphodita.yml to path to the downloaded MorphoDiTa ZIP file.
  3. Download and extract taggers (language models). Put them into a single directory.
  4. Set MORPHODITA_TAGGERS in docker-compose-morphodita.yml to path to the taggers' directory.
  5. Run docker compose -f docker-compose-morphodita.yml up -d --build to build and start Annotace wih MorphoDiTa.

License

Annotace is licensed under GPL v3.0, Spark and MorphoDiTa are distributed under their respective licenses.