CrewScout is a system for finding expert teams in accomplishing tasks. The underpinning concept of the system is skyline teams or skyline groups, which we introduced in a CIKM12 paper and a TKDE paper.
Consider a set D of n experts t1, ..., tn, modeled by m numeric attributes A1, ..., Am that represent their skills and expertise. Any subset of k experts form a k-expert team. CrewScout finds, for a given k, all k-expert skyline teams, i.e., k-expert teams that are not dominated by any other k-expert teams. It further assists users in choosing among the skyline teams. The notion of dominance between teams is analogous to the dominance relation between tuples in skyline analysis. CrewScout calculates for each team an aggregate vector of its experts’ individual vectors. CrewScout provides efficient algorithms for four commonly used aggregate functions—AVG (i.e, SUM, since we only compare teams with equal size), MIN and MAX. A team G1 dominates another team G2 (denoted G1 ≻ G2), if and only if the aggregate value of G1 on every attribute is better than or equal to the corresponding value of G2 and G1 has better value on at least one attribute.
The need for finding expert teams prevails in several application areas, including question answering, crowdsourcing, panel selection, project team formation, and so on. This is illustrated by the following motivating examples.
An attractive characteristic of a skyline team is that no other team of equal size can dominate it. In contrast, given a non-skyline team, there is always a better skyline team. This property distinguishes CrewScout from other team recommendation techniques. The skyline teams consist of the teams that are worth recommending. They become the input to further manual or automated post-processing that eventually finds one team. Admittedly, determining the “best” team is a complex task that may involve more factors than what skyline teams can capture—e.g., which experts are available for a task, whether they have good relationship to work together, and so on. The post-processing is thus crucial. Examples of such post-processing include eye-balling the skyline teams, filtering and ranking them by user preferences, and browsing and visualization of the skyline teams. Particularly, CrewScout provides an interactive tool to assist a human user in exploring and choosing skyline teams.
This material is based upon work partially supported by the National Science Foundation Grants 1018865 and 1117369, 2011 and 2012 HP Labs Innovation Research Awards, and the National Natural Science Foundation of China Grant 61370019. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding agencies.