Cover of Natural Language Processing Across Turkic Languages: Methods and Applications, by Sherzod Hakimov
Free online book · Version 0.1.0

Natural Language Processing Across Turkic Languages

Methods and Applications

How to build and evaluate language technology for a family whose languages share much of their grammar but differ widely in data, tools and benchmarks. The book focuses on what is specific to Turkic: agglutinative morphology and vowel harmony, three major scripts, transfer between related languages, and building corpora, treebanks and benchmarks where few exist.

17 chapters 4 appendices Exercises in every chapter TurkicNLP practicals CC BY-SA 4.0
Start reading → Table of contents Corrections
❓

Three recurring questions

How does morphology affect a task? What changes when the script changes? Which knowledge transfers between related languages? Every chapter returns to them.

🧪

Linguistics plus practice

Each chapter pairs linguistic analysis with a hands-on section, most built on TurkicNLP 0.5.0. Exercises range from Unicode case-folding to training and evaluating models.

🎓

Who it's for

NLP researchers and practitioners, graduate students, and members of Turkic language communities. The code assumes intermediate Python and basic machine learning. Linguistic concepts are introduced as needed.

📚

Companion texts

General NLP is covered briefly, with Jurafsky & Martin's Speech and Language Processing as the companion. Johanson's Turkic is the linguistic reference throughout.

Table of contents

From the language family to the resource frontier

Every chapter opens straight in the online book. Start with the Preface or the publication details.

Chapters 1–2

The Language Family

  1. 1The Turkic World: Languages, Speakers, and the Resource Reality
  2. 2The Computational Phenotype of Turkic Languages
Chapters 3–4

Writing Systems and Corpora

  1. 3Scripts, Encoding, and Orthographic Engineering
  2. 4Building Turkic Corpora at Scale
Chapters 5–7

Morphology

  1. 5Morphological Structure: A Linguistic Foundation
  2. 6Finite-State Morphology: The Rule-Based Approach
  3. 7Neural Morphology and the Tokenisation Problem
Chapter 8

Syntax

  1. 8Syntactic Structure and Universal Dependencies
Chapters 9–11

Applications

  1. 9Sequence Labelling: POS Tagging, NER, and Chunking
  2. 10Machine Translation: Within and Beyond the Family
  3. 11Speech Technologies
Chapters 12–15

Language Models

  1. 12Pretrained Language Models for Turkic
  2. 13Meaning: Embeddings, Semantic Resources, and Retrieval
  3. 14Generation, Dialogue, and Interactive Evaluation
  4. 15Evaluation, Benchmarks, and Responsible Deployment
Chapters 16–17

Cross-Lingual Methods and the Resource Frontier

  1. 16Cross-Lingual Transfer within the Turkic Family
  2. 17Endangered Languages, Digital Preservation, and Ethics
Back matter

Appendices & References

  1. AMathematical Foundations
  2. BComputational Resources and Tools
  3. CTurkic Language Reference
  4. DGlossary
  5. §References
Built with the toolkit

Run every practical yourself

The practical sections use TurkicNLP: transliteration in Chapter 3, Apertium transducers in Chapter 6, neural analysers and the morpheme tokenizer in Chapter 7, UD parsing in Chapter 8, tagging and NER (including the LLM backend) in Chapter 9, NLLB in Chapter 10 and Omnilingual ASR in Chapter 11. Install it once and follow along.

Explore the toolkit → Appendix B: tools
$ pip install "turkicnlp[all]"
import turkicnlp

nlp = turkicnlp.Pipeline("kaz")
doc = nlp("Мен мектепке бардым.")
print(doc.to_conllu())
Corrections & suggestions

Found a mistake? Please open a pull request

The book's source lives at github.com/turkic-nlp/book. For typos, factual errors, broken code or examples in your language, please make a pull request there. For larger suggestions, open an issue first so we can discuss it.

Make a pull request → Open an issue
Cite the book

Citation

Sherzod Hakimov. Natural Language Processing Across Turkic Languages: Methods and Applications. Version 0.1.0, 2026. DOI: 10.5281/zenodo.23324599

@book{hakimov2026turkicnlpbook,
  title   = {Natural Language Processing Across
             Turkic Languages: Methods and Applications},
  author  = {Sherzod Hakimov},
  year    = {2026},
  version = {0.1.0},
  doi     = {10.5281/zenodo.23324599},
  url     = {https://github.com/turkic-nlp/book}
}

License

Unless otherwise noted, the text and figures are licensed under CC BY-SA 4.0.

Program code in the book's repository (code/, scripts/, filters/, and the Python and notebook code in exercises/, experiments/ and tables/) is licensed under Apache 2.0.

AI tools helped with drafting, editing, code review, bibliography checks and the cover artwork. The author reviewed, corrected and checked all AI-assisted material by hand. Details are in the Preface.