2015 Research Projects and Yahoo Fellows Announced for Project InMind

NEWS
Jan 15, 2015

image

On December 15, Yahoo held it's inaugural Fellows Day for CMU students and faculty as part of Project InMind

By Don McGillen True innovation is a result of thoughtful, thorough, and transformative research. And true innovation was the goal we set for ourselves along with Carnegie Mellon University’s School of Computer Science when we announced our joint $10 million, five-year collaboration called Project InMind in February 2014. One year later we are pleased to be well on our way. In 2015, six new Yahoo Fellows under the advisement of CMU faculty and in collaboration with Yahoo Labs research scientists are joining Project InMind. The team will apply its expertise and creativity to an industry-first mobile toolkit that allows for each member to easily experiment with Yahoo’s real-time data services, letting them test new ways that machine learning at scale and interface technologies can improve personalized user experiences. While working on their individual projects (listed below), the six Fellows will research, prototype, and experiment in vivo with a persistent mobile software agent that participates in the user’s daily life, through natural language dialogs enhanced by other interaction modalities like gaze, touch, and gesture. The software agent will provide intuitive, machine-learned access to news, finance data, weather, sports, Yahoo Answers, and other Internet information sources. The competition for funding the Fellows and their novel projects was fierce. CMU faculty co-directors Tom Mitchell and Justine Cassell chose from over forty high-quality proposals. Each InMind grant provides funding for the Primary Investigator from the CMU faculty as well as full support for a PhD student (Yahoo Fellow). We are extremely proud to announce the following six new InMind Yahoo Fellows, Primary Investigators, and their individual projects: CMU Faculty: Norman Sadeh Yahoo Fellow: Bin Liu Project: An InMind personalized privacy assistant Empower InMind users to effectively and efficiently control their privacy through the development of a personalized privacy assistant. This assistant will be capable of incrementally learning user privacy preferences, semi-automatically configuring many privacy settings, and very selectively determining when to interact with users to refine existing models and/or nudge them to review prior decisions. CMU Faculty: Eric Nyberg Yahoo Fellow: Zi Yang Project: Automatic question answering for community question answering (CQA) sites Address the critical research question of how to synthesize and pull together information nuggets combined in multiple answers to previously resolved questions in order to generate a high-quality comprehensive answer for a new arriving question. Automatic question answering could be used to improve the experience of both the original askers as well as subsequent searchers seeking information related to the question -- and serve to augment the human-powered CQA sites. CMU Faculty: Anind Dey, Jennifer Mankoff Yahoo Fellow: Nikola Banovic Project: Computing using routines and behaviors By modeling these routines, a cognitive assistant can compare successful routines to unsuccessful routines (e.g. students who have higher GPAs, sleep more, or have a better work/life balance). On this basis, a system can encourage better routines/behaviors, helping transition users to have behaviors similar to more successful students, provide decision support tools for helping students decide how to make use of available time based on future impact, and automatically provide appropriate information or perform services. We propose to augment the current InMind portfolio of projects by incorporating information about routines into the profile generation and contextual aggregation projects, to support healthier and more effective lifestyles for students. CMU Faculty: Robert Frederking Yahoo Fellow: Justin Chiu Project: Email understanding component The email understanding component (along with others) was crucial to RADAR's success. A similar component will clearly be necessary to achieve InMind's long-term goals. In the shorter term, one clear initial use for the email understanding component will be to help inform customization of the Yahoo News app to user interests. CMU Faculty: Louis-Philippe Morency Yahoo Fellow: Amir Ali Bagherzadeh Project: Modeling nonverbal dynamics By sensing and interpreting nonverbal communicative behaviors in real-time we will enable a large array of innovative application for social agents on mobile platforms. The recent advance of sensing in individual modalities (e.g., computer vision, speech analysis, natural language analysis) are creating a great opportunity for mobile devices to better infer the cognitive and emotional state of the user. Our proposed project will address the crucial challenge of fusing these different modalities in a context-sensitive approach where the valuable information already present on the mobile device (e.g., upcoming appointments, social connections) can be leveraged during the multimodal integration. It will bring mobile technologies one step closer to a more natural interaction with strong sense of rapport and engagement. CMU Faculty: Alexander Rudnicky Yahoo Fellow: Ming Sun Project: An intelligent assistant that learns about and supports task-level activity In the first phase of the project, we want to enable the agent to decide when the app-of-interest (e.g., Yahoo News) should be activated, without being explicitly requested by the user. Our measure of success would be user acceptance and duration of use for this functionality. Subsequently we intend to design and implement a task-level agent that interacts with users by spoken language and provides assistance at that level.