Information Processing & Management最新文献

Purpose

The serious information overload problem of MOOCs videos decreases the learning efficiency of the students and the utilization rate of the videos. There are two problems worthy of attention for the matrix factorization (MF)-based video learning resource recommender systems. Those methods suffer from the sparsity problem of the user-item rating matrix, while side information about user and item is seldom used to guide the learning procedure of the MF.

Method

To address those two problems, we proposed a new MOOCs video resource recommender LSMFERLI based on Low-rank and Sparse Matrix Factorization (LSMF) with the guidance of the inter-Entity Relations and intra-entity Latent Information of the students and videos. Firstly, we construct the inter-entity relation matrices and intra-entity latent preference matrix for the students. Secondly, we construct the inter-entity relation matrices and intra-entity affinity matrix for the videos. Lastly, with the guidance of the inter-entity relation and intra-entity affinity matrices of the students and videos, the student-video rating matrix is factorized into a low-rank matrix and a sparse matrix by the alternative iteration optimization scheme.

Conclusions

Experimental results on dataset MOOCcube indicate that LSMFERLI outperforms 7 state-of-the-art methods in terms of the HR@K and NDCG@K(K = 5,10,15) indicators increased by an average of 20.6 % and 21.0 %, respectively.

Identifying scientific disruption is consistently recognized as challenging, and more so is to predict it. We suggest that better predictions are hindered by the inability to integrate multidimensional information and the limited scalability of existing methods. This paper develops a framework based on graph signal processing (GSP) to predict scientific disruption, achieving an average AUC of about 80 % on benchmark datasets, surpassing the performance of prior methods by 13.6 % on average. The framework is unified, adaptable to any type of information, and scalable, with the potential for further enhancements using technologies from GSP. The intuition of this framework is: scientific disruption is characterized by leading to dramatic changes in scientific evolution, which is recognized as a complex system represented by a graph, and GSP is a technique that specializes in analyzing data on graph structures; thus, we argue that GSP is well-suited for modeling scientific evolution and predicting disruption. Based on this proposed framework, we proceed with disruption predictions. The content, context, and (citation) structure information is respectively defined as graph signals. The total variations of these graph signals, which measure the evolutionary amplitude, are the main predictors. To illustrate the unity and scalability of our framework, altmetrics data (online mentions of the paper) that seldom considered previously is defined as graph signal, and another indicator, the dispersion entropy of graph signal (measuring chaos of scientific evolution), is used for predicting respectively. Our framework also provides advantages of interpretability for a better understanding on scientific disruption. The analysis indicates that the scientific disruption not only results in dramatic changes in the knowledge content, but also in context (e.g., journals and authors), and will lead to chaos in subsequent evolution. At last, several practical future directions for disruption predictions based on the framework are proposed.

Topic evolution has been studied extensively in the field of the science of science. This study first analyzes topic evolution pattern from topics’ semantic consistency in the semantic vector space, and explore its possible causes. Specifically, we extract papers in the computer science field from Microsoft Academic Graph as our dataset. We propose a novel method for encoding a topic with numerous Contextualized Word Embeddings (CWE), in which the title and abstract fields of papers studying the topic is taken as its context. Subsequently, we employ three geometric metrics to analyze topics’ semantic consistency over time, from which the influence of the anisotropy of CWE is excluded. The K-Means clustering algorithm is employed to identify four general evolution patterns of semantic consistency, that is, semantic consistency increases (IM), decreases (DM), increases first and then decreases (Inverted U-shape), and decreases first and then increases (U-shape). We also find that research methods tend to show DM and U-shape, but research questions tend to be IM and Inverted U-shape. Finally, we further utilize the regression analysis to explore whether and, if so, how a series of key features of a topic affect its semantic consistency. Importantly, semantic consistency of a topic varies inversely with the semantic similarity between the topic and other topics. Overall, this study sheds light on the evolution law of topics, and helps researchers to understand these patterns from a geometric perspective.

Rank aggregation is a vital tool in facilitating decision-making processes that consider multiple criteria or attributes. While in many applications, the available ranked lists are often limited and quite partial for various reasons. This scarcity of ranking information presents a significant challenge to rank aggregation effectiveness. To address this problem of rank aggregation with limited information, in this study, on the basis of networked representation of ranking information, we employ the link prediction technology to mine potential ranking information. It aims to optimize the aggregation process, and maximize the aggregation effectiveness using available limited information. Experimental results indicate that our proposed approach can significantly enhance the aggregation effectiveness of existing rank aggregation methods, such as Borda’s method, competition graph method and Markov chain method. Our work provides a new way to solve rank aggregation problem with limited information and develops a new research paradigm for future rank aggregation studies from the perspective of network science.

Learning the preferences of consumers, providers, and system stakeholders is a challenging problem in the Multi-Stakeholder Recommendation System (MSRS). Existing MSRS methods lack the ability to generate equitable recommendations and investigate implicit relationships between stakeholders and items. Hence, this study addresses this issue by proposing a multi-stakeholder preference learning-based recommendation model that exploits information from multiple views to evaluate stakeholders' preferences. The proposed model learns consumer preferences using users' ratings and reviews of an item, provider preferences with provider utility, and provider-item interaction. Furthermore, the proposed model learns the system-level preference of promoting long-tail items through the probabilistic evaluation of stakeholders' interest in popular and unpopular items. Finally, this study develops a multi-stakeholder, multi-view deep neural network model to aggregate stakeholders' preferences and deliver equitable recommendations. This work utilizes benchmark Movie Lens (ML) 25M, ML-100K, ML-1M, and TripAdvisor datasets to validate and compare the proposed model's performance with other baseline methods using standard evaluation metrics for each stakeholder. Examining the precision metrics, the proposed model attains the minimum enhancement of 7.91%, 18.24%, 10.72%, and 20.12% across the ML-25M, ML-100K, ML-1M, and TripAdvisor datasets. Further concerning the exposure, hit, and reach metrics, the model exhibits a substantial minimum improvement of 19.12%, 14.73%, 5.37%, and 28.46% over the ML-25M, ML-100K, ML-1M, and TripAdvisor datasets. Finally, the proposed model excels in promoting long-tail items and enhancing the cumulative utility gain of the stakeholders, surpassing the baseline methods.

Influence maximization (IM) is a key issue in network science. However, previous research on IM has previously explored binary interaction relationship in ordinary graphs, with little consideration for higher-order interaction that are more practical in hypergraphs, especially weighted hypergraphs. Therefore, this study focuses on solving the IM problem in weighted hypergraphs. Firstly, we adopt a novel and more reasonable dissemination model, namely the adaptive dissemination (AD), and incorporate it into weighted hypergraphs. Next, a computing expectation-based influence evaluation method is proposed to accurately obtain the expected influence in one-hop area (EIOA) of the seed node set. Meanwhile, three search algorithms are designed using the EIOA to effectively solve the initial seed set. Then, multi-level experiments are conducted to compare the proposed algorithms with other six advanced algorithms in eight weighted hypergraph datasets from the real world. The experimental results are visually analyzed and two nonparametric test processes are applied to verify the significant advantages of the proposed algorithms. Finally, the impact of different factors such as seed set correlation, model parameter setting, and weight attribute on dissemination is explored, and the efficiency and robustness of these algorithms are further validated.

B-mode ultrasound (BUS) mainly reflects the tissue structural, morphological, and echo characteristics of liver tumors, and contrast-enhanced ultrasound (CEUS) offers supplementary information on the dynamic blood perfusion pattern to promote diagnostic accuracy. Transfer learning (TL) is capable of improving the performance of BUS-based computer-aided diagnosis (CAD) for liver cancer by transferring information from CEUS. However, most multi-view TL algorithms cannot fully capture the view-common together with the view-unique information of three CEUS phase images in the source domain to further promote knowledge transfer. To this end, a multi-view disentanglement-based bidirectional generalized distillation (MD-BGD) algorithm is proposed to explore and learn more potential knowledge from three typical CEUS phase images for multi-view transfer. MD-BGD consists of the multi-view feature disentanglement module and the bidirectional distillation module. The former explores more potential and transferable privileged information by disentangling three CEUS phase image features in the source domain into view-common and view-unique components. The latter develops a bidirectional generalized distillation algorithm to enhance the multi-view knowledge transfer between the source and the target domains, guided by shared labels. Therefore, the BUS-based CAD model is significantly improved by our proposed MD-BGD. MD-BGD is evaluated on the bi-modal ultrasound imaging dataset. It gains the best results of 90.75±2.20 %, 89.50±3.49 %, and 91.89±3.78 %, on accuracy, sensitivity, and specificity, respectively. These results indicate the effectiveness of MD-BGD in the diagnosis of liver cancer.

Existing personalized federated learning frameworks fail to significantly improve the personalization of user preference learning in next Point-Of-Interest (POI) recommendations, causing notable performance deficits. These frameworks do not fully consider crucial factors such as: (1) how to thoroughly explore spatial–temporal relationships within user trajectories to deeply understand personalized behavior patterns, and (2) the neglect of collaborative signals among users with similar spatio-temporal distributions, which results in the loss of valuable shared information. To tackle these challenges, this paper introduces the Spatio-temporal Consistency Federated Learning (SCFL) framework, which capitalizes on the spatio-temporal consistency of trajectories to boost the personalized performance of POI recommendation models in FL. Specifically, we have developed the trajectory optimization module SCA for clients in isolation to extract deeper behavioral patterns from the spatio-temporal distribution of sparse trajectories. Additionally, we present a hierarchical aggregation strategy based on distribution consistency, utilizing intermediate entities called Edges to aggregate similar users, thereby enhancing the model’s learning of shared information. Experimental validation across three real-world datasets (NYC, TKY and Gowalla) and two models (SASRec and SSEPT) with six scalability settings shows that SCFL substantially outperforms eight strong baselines. In six experimental configurations, SCFL achieves a personalized performance improvement of 10.65% over the best baselines. Additional experiments have also validated the superiority of SCFL from various perspectives.

Link prediction is a critical task within the realm of graph machine learning. While recent advancements mainly emphasize node representation learning, the rich information encapsulated within edges, proven advantageous in various graph-related tasks, has been somewhat overlooked. To bridge the gap, this paper explores the potential of incorporating edge representation learning for link prediction and identifies three inherent challenges associated with this approach. We introduce the Edge Contrastive Learning for Link Prediction (ECLiP) framework to tackle these challenges. ECLiP integrates edge information into node representations through edge-level contrastive learning, with a distinctive perspective on treating edges, rather than nodes, as the units of instance discrimination. We first illustrate the implementation of this framework using an established edge representation learning method. However, it incurs significant additional training overhead when the number of edges is huge. To mitigate this issue, we present a computationally efficient variant employing a multi-layer perceptron (MLP) for direct edge representation learning. Conducting rigorous experiments across eight distinct datasets with node counts spanning from 2k to 235k, we demonstrate a noteworthy improvement of over 10% on certain datasets, validating the efficacy of our proposed methodology.

Continuous sign language recognition (CSLR) is essential for the social participation of deaf individuals. The structural information of sign language motion units plays a crucial role in semantic representation. However, most existing CSLR methods treat motion units as a whole appearance in the video sequence, neglecting the exploitation and explanation of structural information in the models. This paper proposes a Structure-Aware Graph Convolutional Neural Network (SA-GNN) model for CSLR. This model constructs a spatial–temporal scene graph, explicitly capturing motion units’ spatial structure and temporal variation. Furthermore, to effectively train the SA-GNN, we propose an adaptive bootstrap strategy that enhances weak supervision using dense pseudo labels. This strategy incorporates a confidence cross-entropy loss to adjust the distribution of pseudo labels adaptively. Extensive experiments validate the effectiveness of our proposed method, achieving competitive results on popular CSLR datasets.