Google Chrome's search bar is now powered by machine learning to offer better suggestions

In an official blog, Google announced a new update for the Chrome browser that brings some new changes to the search bar. According to Google, the new update should help the address bar also known as Omnibox, to provide 'more precise and relevant' web page suggestions.

The Mountain View-based firm said that with the latest Chrome update (M124), Google has integrated machine learning models into the omnibox or search bar. Machine learning will help Chrome provide accurate suggestions aligned with what you are searching for.

Google says that the browser previously relied on 'hand-built and hand-turned formulas' to suggest search results. However, the main problem with this was that these weren't flexible enough to bring improvements to or be adopted in new scenarios.

With the new machine learning models infused in the Chrome search bar, Google can 'collect fresher signals, re-train, evaluate, and deploy new models' as time passes. Since the number one answer when asked for ideas to improve the omnibox was to 'improve the scoring system', the adoption of machine learning in the search bar is a big deal because, as noted by Google, "the scoring system went largely untouched for a long time."

According to Google, the machine learning model in the Chrome omnibox will take into account your previous actions over a URL when suggesting a web page. Meaning, that if you navigated away from a web page in the last few seconds or minutes, the machine learning model will rate that web page a lower score based on its understanding that it was not the site that you were looking for.

Going forward, Google believes that this new machine learning model will open up "many new possibilities to improve the user experience by potentially incorporating new signals, like differentiating between times of the day to improve relevance."

Additionally, Google says that the relevance scoring system should change with time, and thanks to the new scoring system, it can now "simply collect fresher signals, re-train, evaluate, and deploy new models periodically over time" for better results.