Why it's important: Google is adding machine learning features to the address bar of its popular web browser. This unassuming text field known as the “omnibox” because it doubles as both a URL input field and a search box is about to get a major upgrade.
Omnibox has evolved far beyond its humble beginnings as a place to enter website addresses. You can now leverage Google's vast search power to handle any type of query or task. However, the recommendations and results it surfaces are driven by a relatively rigid “set of hand-built and hand-tuned formulas.” That's all about to change.
In a recent post on the Chromium blog, Justin Donnelly, engineering lead for Omnibox in Chrome, said his team is adapting machine learning models to significantly improve Omnibox's “relevance scoring” capabilities. He made it clear that he is working hard. In other words, Omnibox will now better understand the context behind your queries and provide more useful suggestions tailored to your needs.
Donnelly says that when he surveyed his colleagues about ways to improve the Omnibox experience, improving the scoring system was at the top of their wish list. Although current rule-based approaches are effective for a vast number of cases, they lack flexibility and are difficult to adapt to new scenarios organically. Introducing machine learning.
New AI models can generate more nuanced relevance scores by analyzing vast datasets on historical data points such as user interactions, browsing patterns, and frequency of visits to certain sites. For example, we've found that if you quickly leave a webpage, suggestions for that URL will be demoted because it's likely not what you were looking for.
With smarter Omnibox on Windows, Mac, and ChromeOS, suggestions are continually improved and customized based on your changing interests and habits. Donnelly's team also plans to consider incorporating time-of-day awareness, models specific to different user groups such as mobile and enterprise, and other contexts for his signals.
Of course, enabling this kind of deep personalization requires handing over more personal browsing data to Google's machine learning models. How comfortable you are with this trade-off is a personal decision.
Google is gradually rolling out these Omnibox improvements through recent Chrome updates, and machine learning models will start working in the wild starting with version M124, which is expected to be released in the coming months. Although not mentioned in the blog post, it's safe to assume that the update will trickle down to mobile eventually.
