Simply Whistling Or Humming Google Can Find The Title Of The Song

Google illustrations (pixabay)

JAKARTA - Google is rolling out several updates and new features that can make browsing easier to use. One of them is the Hum to Search feature. This is useful when the user tries to find a song or lyrics by just whistling.

To quote Engadget , Friday, October 16, in this case, Google uses machine learning techniques to try to identify whistles or tones. The new feature is available today on iOS and Android with Google Assistant.

How to use it is easy, users just ask Google "What song" or tap the search button for a song, then whistle or sing a piece of lyrics that is remembered for 10 to 15 seconds. Then, Google will show the results based on how big the match of the song is, after which the user can tap the result to listen to it.

This feature is known to be similar to that of Shazam or Soundhound for song identification. Based on Google's narrative, this feature works by using its machine learning model to convert the audio into a number-based sequence representing the melody of the song, then the notes can be compared to existing songs.

The company says they train these models from a variety of sources, including humans singing, whistling or humming, as well as studio recordings. Google also removed instruments and vocal qualities to focus solely on that numerical order.

Hence, the Hum to Search feature should function properly when users search for the desired song. The feature is currently available in English on iOS, and about 20 languages on Android. In the future, the tech giant will add more languages to the feature.

Hum to Search feature (doc. Google)

Other, updated search features also include spelling and general search queries. Includes a new spelling algorithm that uses deep neural networks. Google claims to have significantly improved its ability to decipher misspellings.

"In fact, this single change makes for a bigger improvement to spelling than all of our improvements over the past five years. The new spelling algorithm helps us understand the context of misspelled words, so we can help you find the right result, all in less than 3 years. milliseconds, "says Google .

Neural networks have also been applied to understand subtopics around or of an interest, which Google says will help provide greater diversity of content when users perform broad searches.

"For example, if you search for sports equipment at home, we can now understand the relevant subtopics, such as equipment within a budget, premium options, or small space ideas, and show you more diverse content on search results pages," said Google. .

It should be noted, this feature will begin rolling out at the end of the year. Google added that its BERT language comprehension system is now used in almost every query in English, which is said to have helped users receive higher quality results for questions.