Google has made a significant update to Chrome’s address bar, called Omnibox, incorporating machine learning models to provide more relevant and personalized suggestions.
With the release of Chrome 124 on desktop and ChromeOS platforms, Omnibox has replaced its old rules-based scoring system with a smarter approach based on machine learning. This opens the way to best URL and query predictions tailored to each user’s browsing habits.
Before this update, Chrome relied on “ a set of formulas constructed and adjusted by hand » to sort and surface suggestions in the Omnibox as users type. However, Google acknowledges that this rigid rules-based system “ remained largely intact for a long time » and that he had difficulty adapting to new scenarios.
Also read – Google Chrome becomes paid, a Premium subscription is available
Google adds machine learning to its address bar
By moving to an ML model, Google can now integrate a wider range of contextual signals to more accurately determine which suggestions are most relevant in a given situation. The company cites as an example the “ time since last navigation », the new model reducing the relevance scores for sites immediately revisited, taking into account the user’s likely intention to search for something else.
Google also highlights the benefits inherent in deploying machine learning, which allows “ collect fresher signals, retrain, evaluate and deploy new models periodically over time » in order to continually improve and refine the predictive capabilities of the Omnibox.
The current rollout focuses only on desktop, but Google is already exploring ways to further optimize machine learning models for other use cases. These could include specialized versions for mobile devices, enterprise environments, educational institutions and more, each with their own signals and data models to integrate. It now remains to be seen whether or not the new address bar will be smarter than the previous version. At Microsoft, Edge has a second address bar that works differently.