News · October 25, 2019
Google brings BERT language models into Search
Google is using a transformer language model to interpret search queries, one of its biggest ranking changes in years. It shows these models are now cheap enough to run at web scale.
What happened
- Google said BERT models now help it understand about one in ten English-language searches in the US, with more languages and locales to follow.
- BERT is also being used for featured snippets in the two dozen countries where they are available, and Google said it saw notable improvements in languages such as Korean, Hindi and Portuguese.
- According to Google, it is the first time the company has used its latest Cloud TPUs to serve search results, because some BERT models are too complex for traditional hardware.
- TechCrunch reported that the update works best for longer, conversational queries and that Google had open-sourced BERT's code and pre-trained models in 2018.
Why it matters
For businesses, the message is that the same open-source technique is available to everyone. Pre-trained language models can improve site search, support ticket routing and intent detection in your own products without training from scratch.
For marketing leads, there is no trick to optimize for BERT; it rewards pages that answer natural, specific questions clearly. Watch search traffic for long-tail queries over the coming weeks rather than reacting to day-to-day moves.
Sources
- Google Blog: Understanding searches better than ever before
- TechCrunch: Google brings in BERT to improve its search results
This post is CMM’s own summary and commentary on publicly reported news. Facts are attributed to the sources linked above; please read them for full details. CMM is not affiliated with these publishers or with the companies mentioned, and trademarks belong to their owners. Prepared with AI assistance. Corrections or requests: [email protected]. See our editorial policy.
Want help with this?
Tell us what you need, and we will suggest the right team model on a short call.
Talk to us