The Jacobs Technion-Cornell Institute at Cornell Tech and AOL announced April 20 a research technology called “Immersive Recommendations,” a concept where a user opts in to a tool that translates personal digital traces from one platform into content recommendations in another.
The new technique was developed by Cornell Tech researchers to address “the cold-start problem” – how to engage users with relevant content when they first start using a platform.
As an example, could a service like Netflix suggest better movies for a first-time user if it tapped into his or her Twitter data? Could Meetup.com use your Medium posts to tailor events for you? Researchers will present a paper on the new technique at the 25th International World Wide Web Conference in Montreal, Canada.
The technology is the first to emerge from the Connected Experiences Lab, an endeavor of the Jacobs Technion-Cornell Institute at Cornell Tech in collaboration with AOL.
Traditional recommendation systems are reaching diminishing returns and do not address the cold start problem. At the same time, content consumption is constantly on the rise. The Immersive Recommendations algorithm builds preference profiles on users’ behalf – with the user’s authorization – based on their multi-channel online activities. The user-centric model is used to tailor online services from events to travel and restaurants. This method is likely to benefit not only users, but also startup and local services that do not already have rich, detailed information about user activity.
“People’s digital traces provide a wealth of information about preferences and behavior. Users should be able to unlock their personal data for many purposes, including curating content based on their interests,” said Deborah Estrin, professor of computer science at Cornell Tech.
“We're excited to continue working with AOL on this effort through the Connected Experiences Lab, which is setting a new standard for how industry and academia can work together to advance and commercialize new research,” said Mor Naaman, associate professor of information science at the Jacobs Technion-Cornell Institute and director of the Connected Experiences Lab.“The idea for this research came about when the team at Cornell Tech first brainstormed potential AOL-relevant research directions. In a short time, with support from AOL, we have implemented and built this technology, leading to AOL exploring use of these ideas in product.”
The researchers put together two demonstration systems: Newsfie, which recommends articles from Medium.com and Grouplink, which recommends meetups. From a large-scale offline evaluation and a smaller-scale user study, they found immersive recommendations improved recommendation performances by up to 57 percent and 42 percent for news and meetups, respectively, over the state-of-the-art approaches, and showed clear potential in helping users discover more diverse information.
The technology also provides potential privacy benefits. The current implementation of the tool gives users the control over whether and which services they authorize to share and use their data. Moreover, it allows for a privacy-conscious summary of each user’s interests to be shared with a new service, rather than requiring divulging of detailed user traces as it is in most current recommender systems.
As part of the ongoing collaboration, AOL will continue to support the research through testing and use of opt-in data across its network, and is exploring integration of the tool into a number of product groups.
AOL is particularly interested in this research because of its increased focus on leveraging the mobile footprint it has with Verizon, which acquired AOL in June 2015.
“Capturing signals across as many channels as possible leads to better targeting and syndication of content, especially in mobile settings,” saidWilliam Pence, AOL chief technology officer. “We are interested in how these techniques can be applied to create more relevant and ‘immersive’ ad experiences, specifically in video and in VR [virtual reality].”
AOL also has plans to include quality scores for content and ads, and intends to use user-supplied explicit feedback, like that from this research, to tune recommendation algorithms.
The Jacobs Technion-Cornell Institute at Cornell Tech and AOL founded the Connected Experiences Lab, which started operations in June 2015 to explore and create technologies, including analytical techniques that fuse diverse personal data streams into actionable insights; content personalization that works across delivery platforms; and connectivity tools that deepen and sustain engagement within families and communities. The lab is made up of faculty, postdocs and doctoral students from Cornell Tech, the Jacobs-Technion Cornell Institute and the Technion, along with other collaborators.