Modern Tokenization Techniques for AI & LLMs

Übersicht

This course covers tokenization techniques used in modern AI models, including rule-based methods, subword tokenization (BPE, WordPiece, SentencePiece), and vocabulary optimizations. Learners will implement these methods and understand their impact on NLP model performance.

Lehrplan

• Unit 1: Introduction to Tokenization (Rule-Based Tokenization)
Tokenize Text with NLTK
Sentence Tokenization with NLTK
Extract Monetary Values with Regex
Tokenization Showdown with NLTK and spaCy
• Unit 2: Byte-Pair Encoding (BPE) – Subword Tokenization
Exploring Pre-trained Tokenizers with GPT-2
Using Pre-trained Tokenizers with RoBERTa
Comparing Tokenization with GPT-2 and RoBERTa
• Unit 3: Comparing BPE, WordPiece, and SentencePiece in NLP
WordPiece Tokenization Challenge
Tokenization Techniques in Action
Tokenization Techniques in Action
Tokenization Techniques for Special Texts
• Unit 4: Tokenization and Out-of-Vocabulary (OOV) Handling in NLP
Tokenization Showdown BERT vs GPT2
Multilingual Tokenization Challenge
Multilingual Tokenization and OOV Reduction

---

Fachgebiete

Computer Science