Google yesterday announced a range of AI-powered enhancements to its search engine, including a new algorithm that improves spelling corrections.
The tool uses a deep neural net with 680 million parameters to better understand the context of misspelled words. It runs in 3 milliseconds — faster than one flap of a hummingbird’s wings, according to Google.
The algorithm will be incorporated in Google’s “did you mean” feature by the end of the month. Cathy Edwards, VP of engineering at Google, said it will improve the feature “more in this one day than it has the last five years combined.”
Google is also using AI to improve highly specific searches. These currently don’t always produce the results you need, if the answer isn’t the main focus of a page.
But Google has now developed a way of ranking not just web pages, but also individual passages on the page. This enables the system to zoom onto the relevant section and pick out only the relevant content. Google says it will improve 7% of search queries across all languages when it rolls out next month.
Another update is designed to understand the subtopics around a broader search. For example, when you search for “home exercise equipment,” the system will deliver more specific results, such as budget and smaller gear. This feature will start rolling this out by the end of this year.
Another upcoming Google feature will automatically tag key moments in videos, so you can flick through them like chapters in a book. The company expects 10% of searches to use this by the end of 2020.
Google is also adding a new COVID-19 safety feature to Google Maps. The company says you’ll soon get to see live how busy a place is directly in the map, without having to search for a specific business, which could be handy for anyone trying to avoid crowds during the pandemic.
Get the Neural newsletter
Greetings Humanoids! Did you know we have a newsletter all about AI? You can subscribe to it right here.Follow @neural