10 March 2025 in ACM Transactions on Multimedia Computing, Communications, and Applications
Knowledge-aware video recommendation requires the ability of associating external knowledge to capture high-order connectivities between users and videos. One limitation of existing methods is that they only extract user interests at a granular level of relational paths by modeling high-order connectivities, which are coarse-grained in user interest modeling, failing to identify user–video relations at a finer grained level of semantics. In this article, we investigate the utility of semantic scene graphs in video recommendation scenario, which provide detailed, graph-based annotations of social situations depicted in video clips. We propose a new method named Scene-Centric Multi-Level Interest Miner (SCIMiner) , which explicitly models multi-level user interests. Specifically, we construct user–video graph, knowledge graph, and semantic scene graphs as hierarchical heterogeneous graphs. On top of the hierarchical graph representation, we propose two information aggregation strategies to capture user interests from different levels. Knowledge-aware aggregation scheme extracts coarse-grained user interests by aggregating relational paths in high-order connectivities, while scene-aware aggregation scheme models fine-grained user interests by capturing clip-level semantic commonality of user favorite videos. Furthermore, we adaptively distill complementary information about multi-level user interests extracted by different aggregation schemes and encode them into the representations of users and videos. Empirical results show that SCIMiner significantly outperforms the state-of-the-art methods. Further studies verify the complementary benefits of knowledge graph with semantic scene graphs in video recommendation scenario and the finer grained explainability for predictions.
17 October 2025 in IEEE Transactions on Computational Social Systems
Roll call vote prediction is a research field in quantitative political science that is crucial for understanding the legislative process and assessing the impact of political decisions. Much of the current work incorporates contextual information about legislation and legislators, relying on historical voting data to learn about legislation and preferences of legislators. However, these efforts have paid insufficient attention to characterizing the important role of cosponsors in legislation. In practical scenarios, whether a legislator votes for or against a bill is largely related to the cosponsors. To address the above challenge, in this article, we propose a group enhanced roll call vote prediction model, which considers cosponsors as a group. Specifically, we first reconstruct the historical cosponsorship network to obtain an accurate graph of relationships among legislators, and we utilize a GNN to obtain the preferences of legislators. Then, we propose an individual aggregation method to obtain the representation of groups and we further employ the BERT model to obtain the representation of legislative texts. We further integrate these representations by attention mechanisms to gain the final representation of legislation. It is worth mentioning that in the group representation component, considering the large size of some groups and the significant differences among group members, we apply self-attention mechanism within the group and prune members based on attention scores. Finally, we input the obtained representations of legislation and legislators into an MLP layer to obtain the final voting results of legislators on the legislation. Extensive experiments conducted on real-world datasets demonstrate the competitiveness of our model compared to state-of-the-art baseline methods.
1 January 2026 in IEEE Transactions on Computational Social Systems