ahmetaa/zemberek-nlp

Zemberek‑NLP – Turkish Natural‑Language Processing Toolkit

What it is – A Java library that bundles a collection of NLP components specifically for Turkish. It covers everything from low‑level text handling (tokenization, spelling correction) up to higher‑level tasks such as morphological analysis, named‑entity recognition, language identification and text classification.

Main modules

Module	What it does
Core	Utility collections, hash functions, and helper classes used by the other modules.
Morphology	Turkish morphological parsing, disambiguation and word‑form generation.
Tokenization	Sentence boundary detection and tokenisation tuned for Turkish orthography.
Normalization	Spell‑checking, word‑suggestion and noisy‑text cleaning.
NER	Named‑entity recogniser (model not yet shipped).
Classification	Text classification built on a Java port of Facebook’s fastText.
Language Identification	Quick detection of the language of a given text.
Language Modeling	Compression‑friendly language‑model format and utilities.
Applications	Ready‑to‑run console tools (e.g., interactive morphology analyser).
gRPC Server	Exposes the library over gRPC so it can be called from other languages.
Examples	Small programs showing typical usage patterns.

How to use it

• Maven – Add the Zemberek Maven repository (https://raw.github.com/ahmetaa/maven-repo/master) and declare the desired module(s) as dependencies, e.g. zemberek-morphology version 0.17.1.
• Standalone JAR – Download zemberek-full.jar from the linked Google‑Drive folder. Running java -jar zemberek-full.jar lists the bundled console applications.
• Source – The repository contains a examples module and a separate turkish-nlp-examples project that demonstrates the API.

Current status – The project is in slow maintenance mode; the latest release is 0.17.1 (July 2019). Some components (e.g., NER) lack pretrained models, and the library has limited multi‑threading testing.

License – Apache License 2.0 (permissive, commercial‑friendly).

Who might need it – Researchers, developers, or hobbyists building Turkish‑language applications (chatbots, search, text analytics, etc.) who prefer a pure‑Java solution that can be embedded directly or accessed via gRPC.