Providing proactive support for seeking, sharing, and making sense of information.
Answering the questions you should be asking.
Giving you guidance for the future that you may not know exists.
Information seeking barriers, failures, and solutions. In an effort to understand where people fail in information seeking situations and how to best help them, even when they don't explicitly ask for help, we collected data from a qualitative survey involving 208 real-life examples of information seeking failures. In addition, 10 semi-structured interviews with 10 different participants were analyzed using various theoretical frameworks of tasks, strategies, and barriers.
Generative query recommendations. Sometimes, to help an information seeker, we need to go beyond just incremental changes to their existing expression of information need. This is where we have developed new techniques for generating queries that allow a searcher to make a leap rather than just a step. Given a user’s past queries, we model their past queries as a set of topics and recommended the least explored topic that was still relevant to their search, according to a threshold criterion. We then generate multi-word query terms from this topic, using a skipgram model that could create coherent phrases.
Strategy recommendations. We have conducted research in recommending strategies (a sequence of steps involving queries, documents, and relevant information) for exploratory search tasks. We were able to accurately predict the performance of a user a few steps ahead within their search, and moreover could offer effective recommendations. The recommendations offered became more effective as the recommendations were given later in the search process. They moreover greatly enhanced user performance, according to quantitative simulations and qualitative judgments of the simulations and user performance.
Collaborator recommendation. Some times, to help someone in their information seeking process, we should recommend a person to collaborate with. Using 120 participants in 60 pairs working on an exploratory search task, we built a model for finding suitable collaborators for an individual during different stages of his/her search. This model was then tested on real-life data obtained from Bing search logs and Internet Explorer browser logs. The data consisted of 8,969 search sessions by 8,051 users who were working on an exploratory search task during the overlapping timeframe.
Information Fostering: Being Proactive in Information Seeking. Microsoft Research. Redmond, Washington. March 29, 2017.
Information Fostering: Proactively Complementing Information Seeking. GESIS in Cologne, Germany. June 30, 2016.
Information Fostering: Being Proactive in Information Seeking. University of Texas at Austin. March 4, 2016.
Information Fostering: Proactively Complementing Information Seeking. Guest lecture in PhD 610 (Seminar in Information Science). Rutgers University. October 13, 2016.
Information Fostering: Proactively Supporting Information Seeking. University of Tampere, Finland. October 21, 2016.
Minority Report for Information Seeking: Fixing Things Before They Break in Search. University Technology at Sydney, Australia. October 21, 2015.