Machine Learning Powered Reading App



Product Lead: Yours truly / UX Design: Yours truly, Seph L / UI Design: Seph L / Dev: Luciano L, Shaolie L, Cheng Y

There is a rising demand for English articles in China, with more and more people using apps like CNN and Flipboard. However, the experience of these apps can be pretty awful in China. The content loads slowly and is sometimes inaccessible with a VPN. Even with available content, there is a language and cultural barrier that makes reading the articles quite tough.

To help the Chinese audience deal with these pain points we built Seed. Seed aggregated English sources and provided Chinese readers a speedy way to access these articles, as well as a community to read along with. Seed also allowed users to look up any unknown words or phrases and annotate any passages of interest. Through machine learning, Seed also dynamically analyzed each article for important people, places and objects and offered detailed Chinese wiki definitions to establish a stronger cultural context. 


This application was our major project for around 18 months and as such went multiple iterations and redesigns. But one thing stay constant through these iterations: article browsing experience. The initial issues that we faced was wanting to give the user the simultaneous feeling of focus and variety. We didn’t want the browsing experience to be a table of article titles and thumbnails, as this lacks focus. We also didn’t want to have one article dominating the browsing view, as it would be inefficient in combing through multiple articles. To obtain the optimal balance of both focus and variety, we designed a horizontal stream of article thumbnails in their original ratios. As the user pulled a thumbnail toward center of the screen, it would enlarge and the title of the article would appear. The expanded photo would dominate the screen but allowed space for other thumbnails in the stream reminding the user that there was always more to see, but letting her focus on the article at hand.  We built the stream velocity-sensitive, so if the user swiped short and slow, the browsing experience would be centered and calm but if the user did a big swipe, the thumbnails would fly by allowing for maximum efficiency.