What you need to know
- Google will now use machine learning in Chrome's address bar to provide better, more personalized results.
- The company says it is replacing hand-tuned formulas that were difficult to change with adaptable ML models that can be quickly adjusted and refined over time.
- This change is available in the latest version of Chrome for desktop (version M124).
Google is making internal changes to the latest version of Chrome that aim to improve the web page suggestions that appear in the address bar (also known as omnibox).
In Chrome 124, these suggestions will now be made with the help of machine learning models, which the company says will replace “hand-built and hand-tuned formulas.” Because the address bar leverages ML models, the results should be more accurate and customized for each user.
Justin Donnelly, a Chrome engineering lead working on Omnibox, explains in a blog post that the old scoring system couldn't adapt or change over time. The engineer described it as “inflexible” and said the lack of flexibility “left the scoring system largely untouched for a long time.” So when considering how to improve the Address his bar and its suggestions, the Chrome team thought machine learning was the obvious solution.
ML models can often detect trends and insights beyond the human eye. That also applies to models that power the Omnibox. One of the most obvious changes in address bar behavior with the switch to ML is a change in how the “time since navigation” signal is recognized. Previously, the manual formula gave higher relevance scores to URLs that were recently accessed. However, the ML model discovered that this was not actually what the user was looking for.
“We found that the training data reflected a pattern where users would visit a URL that wasn't what they actually wanted, then immediately return to the Chrome Omnibox and try again,” Donnelly explains. . “In that case, the URL the user just navigated to is almost certainly do not have This second attempt should give you a lower relevance score because that's what they want. ”
Apart from changing the way it scores results by relevance, Google plans to use ML models in the address bar to make web page suggestions “more accurate and relevant to users.” is. Presumably, your browsing habits and other data Google collects will be used to fine-tune Omnibox's behavior to suit your needs. In other words, the way users use the Chrome address bar can be used to retrain ML models that improve over time.
The new address bar is included in Chrome 124 for desktop, but you won't notice any visual differences. In the future, Google hopes to add more signals to consider in its relevance score, such as time of day and environment.
