Published Nov 11, 2022

SDS 626: Subword Tokenization with Byte-Pair Encoding — with @JonKrohnLearns

Jon Krohn delves into the world of tokenization in natural language processing, spotlighting the innovative subword methods with a deep dive into byte-pair encoding's role in enhancing models like BERT and GPT-3.

Episode Highlights

Topics covered

Tokenization Methods
01
Byte-Pair Encoding
02

Popular Clips

Tokenization Techniques
Play Clip

SDS 626: Subword Tokenization with Byte-Pair Encoding — with @JonKrohnLearns​

Topics covered

Popular Clips

Tokenization Techniques

Episode Highlights

Tokenization MethodsJon Krohn explores the intricacies of tokenization in NLP, focusing on word, character, and subword methods. He highlights the strengths and limitations of each approach, emphasizing the power of subword tokenization with byte-pair encoding.

Tokenization Methods

Byte-Pair EncodingJon Krohn explores byte-pair encoding (BPE) and its transformative role in subword tokenization for natural language processing (NLP). He explains how BPE enhances the flexibility and efficiency of NLP models like BERT and GPT-3.

Byte-Pair Encoding

SDS 626: Subword Tokenization with Byte-Pair Encoding — with @JonKrohnLearns