Yandex company blog

Yandex Rolls Out Experimental Social Recommendation Tool – Wonder

Social networks have become the main source of information for many of us. We get news from Twitter and Facebook, learn about new places on Foursquare and Instagram, discover new music on Facebook via Spotify and Rdio. We receive information through our news feeds which change so quickly that we often lose track of what has just caught our attention. To help ourselves retrieve interesting bits from our social networking space, we have built an app for iPhone and iPod Touch that can respond to our questions using the information about our friends’ activity on Facebook, Instagram, Foursquare and Twitter – places they visited, music they listened to, or news they read – and called it Wonder

Wonder supports natural language, which is English so far, and features a voice user interface built using the speech recognition and text-to-speech conversion technologies provided by Nuance Communications, Inc. (NASDAQ:NUAN). Making the most of your social networking has never been easier and more effective:

- If you are looking for a proven sushi place in New York, you can just ask: what sushi restaurants do my friends go to in New York?
- When you are looking for coffee shops in a new area, you can ask: coffee shops nearby.
- If you need to catch up with your friends on a Friday night, just ask: where do my friends party?
- You know your friend John has a good taste for music, ask: what music does John listen to?
- Feel like listening to electronic music, ask: I wonder what electronic music are my friends listening to?
- Want to catch up on news, ask: news shared by my friends.

The information that the user receives in Wonder in response to their questions comes from their friends across a number of social networks. For example, a location-related question brings visual stacks of pictures together with friends’ tips from Facebook, Instagram and Foursquare. The user only has to tap on a stack to unfold the pictures and swipe through them.

When in a crowded place and uncomfortable to speak to their phone in public, the user can type – there is the keyboard button under the activated microphone icon. To facilitate the input, Wonder provides personalized suggestions of friends’ names, cities they visited, place categories, and music genres.

Wonder is integrated with a number of web services (Last.fm, Foursquare’s venue database, iTunes), which allows users to get more information and consume what they have found without leaving the application. For example:

Tap on a restaurant name to see the menu and call to book a table.



Tap on the album title to read the artist’s bio or tap on the album cover to preview a song and buy it on iTunes.

The user can start asking questions in Wonder a few minutes after they have provided access to their social network accounts and the first portion of data has been collected. The more data is collected over time, the better the answers.

Wonder is implemented at Yandex Labs using Yandex’ technologies for natural language understanding and social data collector. The natural language understanding unit translates questions in English into structured database queries. The social data collector is built from scratch, it periodically fetches, cleans and stores social data into a database designed to run structured queries. Read more about the app’s mechanics here.

Wonder is an experiment. Currently, it supports three domains (places, music and news) and only a few types of questions. The app’s scope will be expanded based on users’ feedback. Made by Yandex’s Palo Alto team with the U.S. audiences in mind, the app may become a welcomed and long-awaited tool for the users of social networks in Russia, such as VK, the country’s largest and most popular social networking website. 

Wonder can be downloaded for free from the U.S. Apple App Store. We would love to hear from you - feel free to share your perceptions here

More screenshots of Wonder app are available at Yandex.Fotki photo album.

Visit the app's website at wonder.yandex.com.