Google has detailed a number of new improvements to its Search Included event, which it will make in its core Google search service in the coming weeks and months. The changes are largely focused on the use of new artificial intelligence and machine learning techniques to provide better consumer demand results. Chief among them: a new spell checker that Google promises will help identify even the worst-written queries.
According to Prabhakar Ragavan, head of Google search, 15 percent of Google search queries every day are ones that Google has never seen before, which means the company must constantly work to improve its results.
Part of this is due to poorly written requests. According to Kathy Edwards, vice president at Google, 1 in 10 Google search queries is wrong. Google has long been trying to help with its “you meant” feature, which suggests correct spellings. By the end of the month, it will release a massive update to this feature, which uses a new spelling algorithm powered by a neural network with 680 million parameters. It works less than three milliseconds after each search, and the company promises to offer even better suggestions for misspelled words.
Another new change: Google search can now index individual passages from web pages instead of just the entire web page. For example, if users search for the phrase “how can I determine if my house’s windows are UV glass”, the new algorithm can find a paragraph on the DIY forum to find the answer. According to Edwards, when the algorithm begins to spread next month, it will improve 7 percent of requests in all languages.
Google also uses AI to split the wider search for subtopics to provide better results (say, helping to find homework equipment designed for smaller apartments, instead of simply providing general equipment information. for training).
Finally, the company is also beginning to use computer vision and speech recognition to automatically tag and split videos into parts, an automated version of the existing chapter tools it already offers. For example, cooking videos or sports games can be analyzed and automatically divided into chapters (something Google already offers creators to do by hand), which can then appear in a search. This is a similar effort to the company’s existing work of displaying specific episodes of podcasts in search, rather than simply showing the total feed.