220 likes | 224 Views
Exploring effective collaboration in expertise networks to form skilled teams with efficient communication for task completion.
E N D
Finding a team of Experts in Social Networks Theodoros Lappas UC Riverside Joint work with: Evimaria Terzi (IBM Almaden), Kun Liu (IBM Almaden)
Motivation “ How can I find ateam of experts that cancollaborate effectively in order to complete a given task? ”
Problem • Given a task and a set of experts organized in a network, find a subset of experts that can effectively perform the task • Task: set of required skills • Expert: an individual with a specific skill-set • Network: represents strength of relationships
Expertise networks • Collaboration networks (e.g., DBLP graph, coauthor networks) • Organizational structure of companies • LinkedIn • Geographical (map) of experts
Eleanor Eleanor David Cynthia {graphics,java,python} {graphics,java,python} {graphics} {graphics, java} Bob Alice Alice {python} {algorithms} {algorithms} What makes a team effective for a task? • T = {algorithms, java, graphics, python} Coverage:every required skill in T is included in the skill-set of at leastone team member
Eleanor David Cynthia {graphics,java,python} {graphics} {graphics, java} Bob Alice {python} {algorithms} Is coverage enough? T={algorithms,java,graphics,python} A,E could perform the task if they could communicate A A D A,B,C form an effective group that can communicate B B C C E E Communication:the members of the team must be able to efficiently communicate and work together
Problem definition • Given a task and a social network of individuals G, find the subset (team) of G that can effectivelyperform the given task. • Thesis: Good teams are teams that have the necessary skills and can also communicate effectively
How to measure effective communication? The longest shortest path between any two nodes in the subgraph • Diameter of the subgraph defined by the group members A A D B B C C E E diameter = infty diameter = 1
How to measure effective communication? The total weight of the edges of a tree that spans all the team nodes • MST (Minimum spanning tree) of the subgraph defined by the group members A A D B B C C E E MST = infty MST = 2
Problem definition – v.1.1 • Given a task and a social network G of individuals, find the subset (team) of individuals that can perform the given task and define a subgraph in Gwith the minimum diameter. • Problem is NP-hard
The RarestFirst algorithm T={algorithms,java,graphics,python} {graphics,python,java} {algorithms,graphics} A B A B Skills: algorithms graphics java python E E {algorithms,graphics,java} C D {python,java} {python} αrare= algorithms Srare={Bob, Eleanor} Diameter = 2
The RarestFirst algorithm T={algorithms,java,graphics,python} {graphics,python,java} {algorithms,graphics} A B Skills: algorithms graphics java python E E {algorithms,graphics,java} C C D {python,java} {python} αrare= algorithms Srare={Bob, Eleanor} Diameter = 1 Running time: Quadratic to the number of nodes Approximation factor: 2xOPT
Problem definition – v.1.2 • Given a task and a social network G of individuals, find the subset (team) of individuals that can perform the given task and define a subgraph in Gwith the minimum MST cost. • Problem is NP-hard Best known Approximation factor: O(log3n log k)
The SteinerTree problem • Graph G(V,E) • Set of Required Vertices R • Find G’ subgraph of G such that G’ contains all the required vertices (R) and MST(G’) is minimized Required vertices
The EnhancedSteiner algorithm T={algorithms,java,graphics,python} graphics {graphics,python,java} {algorithms,graphics} A B java algorithms E E {algorithms,graphics,java} D C D python {python,java} {python} MST Cost = 1
Dataset DBLPDataset ( DM, AI, DB, T ) ~6000 authors Skills: keywords appearing in paper titles ~2000 features Social Network: Co-Authorship Graph Tasks: Subsets of keywords with different cardinality
Example teams (I) • S. Brin, L. Page: The anatomy of a large-scale hypertextual Web search engine • Paolo Ferragina, Patrick Valduriez, H. V. Jagadish, Alon Y. Levy, Daniela Florescu, Divesh Srivastava, S. Muthukrishnan • P. Ferragina ,J. Han, H. V.Jagadish, Kevin Chen-Chuan Chang, A. Gulli, S. Muthukrishnan, Laks V. S. Lakshmanan
Example teams (II) • J. Han, J. Pei, Y. Yin: Mining frequent patterns without candidate generation • F. Bonchi • A. Gionis, H. Mannila, R. Motwani
Extensions • Other measures of effective communication • Other practical restrictions • Incorporate ability levels