Speak into the microphone and you’ll be trained on Google.

Google Search is releasing a new feature that allows users to check the pronunciation of unfamiliar words through machine learning.

When searching for a pronunciation guide, the user can speak into the microphone, and Google uses AI to analyze the pronunciation of the word. Then you get feedback on how each syllable matches our expected pronunciation.

For example, if you practice how to say “asterisk”, speech recognition technology analyzes how the user said the word and then recognizes that the last sound is pronounced as ‘rict’ instead of ‘uhsk’. You will receive feedback on how you can improve next time. ”

This feature starts today, but Google says it’s “experimental” and only available on mobile. Google says Spanish pronunciation “will come along,” but the new guide applies only to American English words.

We’re improving word translation and definitions with visual prompts. For example, if you try to translate “naranja” from Spanish to English, you will also see an orange picture with the translated words. If you look at the meaning of the word “seal,” you’ll see mechanical seals, embossed wax pieces, and photographs of semi-aquatic marine mammals. There is no confusion there.

Google says that these photo translations only work on nouns, the words that are best visualized at first in English. However, we plan to expand coverage in the future.

Both features seem to be useful additions to Google’s already impressive language technology, replacing simple search with a more balanced language coach. Machine learning and the data Google can get from people who use this feature will improve.