Preface

Who this book is for

This book grew out of an Applied NLP course at METU (ODTÜ) taught to engineers from other disciplines — chemical, industrial, civil, electrical — and to statistics students. It assumes you are comfortable with probability, linear algebra at the level of Appendix B — Math Refresher, and basic Python. It does not assume data structures, algorithms courses, or any prior machine learning.

Two promises follow from that audience:

1. Statistics first. Every new idea is introduced through something you already trust — estimators, regression, weighted averages — before it gets its deep-learning name.
2. Applied means applied. Each part ends in something you can run, evaluate, and defend in front of your own organization.

How to read it

Parts I and II are sequential and build the vocabulary everything else depends on. From Part III onward the chapters are largely independent: pick the task, the LLM technique, or the production concern you need. Parts V–VII assume Part IV. Interactive widgets run in the browser; executable chapters carry runnable Python (Colab notebooks are not published yet). Exercises are few and deliberately open-ended — they are the course’s homework seeds.

About updates

This is a living book. Substantive changes are recorded in the update log, and the Git history is the full audit trail. Cite a version and include the access date — Citation, License, and Colophon shows how.

Türkçe okuyucular için not

Bu kitabın Türkçe sürümü yayında. Türkçe sürüm bir çeviri eki değil, ayrı bir baskıdır: örnekler mümkün olan her yerde Türkçedir ve bu örneklerdeki sayılar Türkçe veri üzerinde yeniden hesaplanmıştır. Terimlerin iki dildeki karşılıkları için Appendix C — Glossary · Sözlük (EN ↔ TR) bölümüne bakabilirsiniz.

Acknowledgments

To the students at METU who read these chapters as drafts and said which parts did not land; and to the communities behind the open-source tools this book is built on — scikit-learn, gensim, Hugging Face, Quarto, and Observable.

Colophon

Written in Quarto; interactives in Observable JS; source on GitHub under CC BY 4.0 (text) and MIT (code). The full edition record, licenses, and citation entries are in Citation, License, and Colophon.