Google detailed a host of new improvements at its “Search On” event that it will make to its foundational Google search service in the coming weeks and months. The changes are largely focused on using new AI and machine learning techniques to provide better search results for users. Chief among them: a new spell checking tool that Google promises will help identify even the most poorly spelled queries.
According to Prabhakar Raghavan, Google’s head of search, 15 percent of Google search queries each day are ones that Google has never seen before, meaning the company has to constantly work to improve its results.
Part of that is because of poorly spelled queries. According to Cathy Edwards, VP engineering at Google, 1 in 10 search queries on Google are misspelled. Google has long tried to help with its “did you mean” feature that suggests proper spellings. By the end of the month, it’ll be rolling out a massive update to that feature, which uses a new spelling algorithm powered by a neural net with 680 million parameters. It runs in under three milliseconds after each search, and the company promises it’ll offer even better suggestions for misspelled words.
Another new change: Google search can now index individual passages from webpages, instead of just the whole webpage. For example, if users search for the phrase “how can I determine if my house windows are UV glass,” the new algorithm can find a single paragraph on a DIY forum to find an answer. According to Edwards, when the algorithm starts to roll out next month, it’ll improve 7 percent of queries across all languages.
Google is also using AI to divide broader searches into subtopics to help provide better results (say, helping find home exercise equipment designed for smaller apartments versus just providing general workout gear information).
Lastly, the company is also starting to use computer vision and speech recognition to automatically tag and divide videos into parts, an automated version of the existing chapter tools it already provides. Cooking videos, for example, or sports games can be parsed and automatically divided into chapters (something Google already offers to creators to do by hand), which can then be surfaced in search. It’s a similar effort to the company’s existing work in surfacing specific podcast episodes in search, instead of just showing the general feed.