Title: Learning to Form Teams of Skill based Teams of Experts
Year: 2020
Venue: CIKM
Introduction:
Its purpose is to learn feature representations over a set of teams of expert using a variational Bayesian neural architecture.
Main Problem
• This paper focus on finding optimal group of experts, such as group of co-authors which are able to satisfy two main criteria :
They are able to provide maximal coverage for a set of required of skills.
Have effective collaboration history
• Also, the problem is to search for variational distributions of experts and skills in the context of team whose expectations can be efficiently approximated with variational posterior to draw probable teams which more effective than graph based methods.
Input
• Set of required skills to form a team of experts are the given inputs
Output
• The top-𝑘experts with highest probabilities would form the predicted team given the input skills.
Previous Works and their Gaps:
• The previous works are based on the shortest path between experts on the expert collaboration network, and suffer from three major shortcomings:
They are computationally expensive due to the complexity of finding paths on large network structures.
They use a small portion of the entire historical collaboration network to reduce the search space.
They fall short in sparse networks where the majority of the experts have only participated in a few teams in the past
Proposed Method
• Our proposed model is a variational Bayesian neural network (vBnn) with the hidden layer. The top-𝑘 experts with highest probabilities would form the predicted team given the input skills.
Experiments:
• We used DBLP datasets to form skill set S from set of keywords extracted.
Metrics
• We report the average performance using ranking metrics: mean average precision (map), mean reciprocal rank (mrr), normalized discounted cumulative gain (ndcg), and recall for the top-𝑘experts of highest probabilities in the ranked list of predictions.
Gaps of this Work:
• Unable to generalize approach to support for role-based membership of experts in a team given a set of input skills and associated roles
• Only considers two main constraints required skills and collaboration history
Title: Learning to Form Teams of Skill based Teams of Experts Year: 2020 Venue: CIKM
Introduction: Its purpose is to learn feature representations over a set of teams of expert using a variational Bayesian neural architecture.
Main Problem • This paper focus on finding optimal group of experts, such as group of co-authors which are able to satisfy two main criteria :
Input • Set of required skills to form a team of experts are the given inputs
Output • The top-𝑘experts with highest probabilities would form the predicted team given the input skills.
Previous Works and their Gaps: • The previous works are based on the shortest path between experts on the expert collaboration network, and suffer from three major shortcomings:
Proposed Method • Our proposed model is a variational Bayesian neural network (vBnn) with the hidden layer. The top-𝑘 experts with highest probabilities would form the predicted team given the input skills.
Experiments: • We used DBLP datasets to form skill set S from set of keywords extracted.
Metrics • We report the average performance using ranking metrics: mean average precision (map), mean reciprocal rank (mrr), normalized discounted cumulative gain (ndcg), and recall for the top-𝑘experts of highest probabilities in the ranked list of predictions.
Gaps of this Work: • Unable to generalize approach to support for role-based membership of experts in a team given a set of input skills and associated roles • Only considers two main constraints required skills and collaboration history