Lempel-Ziv

• Separates input stream into tokens

• Each token represents the shortest phrase that has not been seen

• Tokens are numbered

• Tokens contain other tokens

• Compression is inefficient at the start

• Better later

• Larger dictionary