The Way to Success: A Multi-level Attentive Embedding Framework for Proposal Teamwork Analysis in Voting-Oriented System

Effective teamwork is crucial for successful proposals, particularly in a competitive voting-oriented system. To increase the chances of success, teams should select collaborators with complementary expertise and broad influence to attract potential supporters. Traditional efforts in quantitative proposal analysis mainly focus on the voting decision process of individual voters from the perspective of social influence, while the teamwork effectiveness of proposals has been largely ignored. To bridge this gap, we propose a novel Multi-level Attentive Embedding Framework (MAEF) to reveal the secret behind a successful proposal from the perspective of teamwork, in which two levels of embedding have been learned to measure the teamwork effects. Specifically, at the individual level, expertise learning and a multi-channel Graph Attention Neural Network are developed for each collaborator to estimate the degree of expertise on different proposal topics and the attractiveness to supporters, respectively. At the team level, we aggregate the individual-level embeddings of collaborators to measure the overall teamwork effectiveness, taking into account the complementarity of expertise and closeness of collaboration. In this phase, the attention mechanism is introduced to distinguish individual contributions. Extensive experiments on a real-world legislative institution dataset clearly validate the effectiveness of our MAEF model compared with several state-of-the-art baseline methods, supporting the hypothesis that effective teamwork indeed improves the predictability of proposal success.