Arnab Nandi, Stelios Paparizos, John Shafer, and Rakesh Agrawal
A typical person has numerous online friends that, according to studies, the person often consults for opinions and advice. However, public broadcasting a question to all friends risks social capital when repeated too often, is not tolerant to topic sensitivity, and can result in no response, as the message is lost in a myriad of status updates. Direct messaging is more personal and avoids these pitfalls, but requires manual selection of friends to contact, which can be time consuming and challenging. A user may have difficulty guessing which of their numerous online friends can provide a high quality and timely response. We demonstrate a working system that addresses these issues by returning an ordered subset of friends predicting (a) near-term availability, (b) willingness to respond and (c) topical knowledge, given a query. The combination of these three aspects are unique to our solution, and all are critical to the problem of obtaining timely and relevant responses. Our system acts as a decision aid — we give insight into why each friend was recommended and let the user decide whom to contact.
|Published in||Proc. 29th International Conference on Data Engineering (ICDE)|