Skip to content
Handbook navigation

Full-access lesson

BERT

The query "2019 brazil traveler to usa need a visa" turns on its smallest word. "To" makes the traveller a Brazilian bound for the United States, and it does so only in the light of the words on both sides of it. Google's ranking missed that and returned pages about U.S. citizens travelling to Brazil. The query was the company's own example on 25 October 2019, when it announced that a year-old research model would now help rank search results and that "BERT will help Search better understand one in 10 searches in the U.S. in English." A model that reads a query once, left to right, can only give "to" the context that came before it.

The complete curriculum

All 12 volumes, companion resources, interviews, and architecture reviews.

Executable engineering practice

Subscriber-only Python, Java, TypeScript, and SQL labs in the isolated runner.

BERT | KnowledgeOS