Information Processing & Management最新文献

{"title":"Meta-PKE: Memory-Enhanced Task-Adaptive Personal Knowledge Extraction in Daily Life","authors":"Yijie Zhong ,&nbsp;Feifan Wu ,&nbsp;Mengying Guo ,&nbsp;Xiaolian Zhang ,&nbsp;Meng Wang ,&nbsp;Haofen Wang","doi":"10.1016/j.ipm.2025.104097","DOIUrl":"10.1016/j.ipm.2025.104097","url":null,"abstract":"<div><div>In this paper, we propose the task of personal knowledge extraction to get structured knowledge from personal data in daily life. The existing information extraction methods struggle to handle this task due to the personal data’s multi-source, fine-grained, dynamic, and personalized nature. They fail to select necessary extraction tasks adaptively, cope with diverse scenarios in daily life, and overlook the assistance of historical personal data for the extraction task. Thus, we propose a novel <strong>M</strong>emory-<strong>E</strong>nhanced <strong>T</strong>ask-<strong>A</strong>daptive <strong>P</strong>ersonal <strong>K</strong>nowledge <strong>E</strong>xtraction method called <em>Meta-PKE</em>. We introduce a task selection module to select the necessary extraction tasks without manual specification according to input personal data. When executing the selected extraction tasks, we record the historical data as the memory and design a memory-enhanced progressive extraction module. Structured personal knowledge is extracted in a coarse-to-fine manner aided by the optimal historical data from a carefully designed memory selection strategy. In addition, we propose a knowledge re-identification module to ensure the completeness of the extracted personal knowledge while avoiding the hallucinations engendered by the large language models. Extensive experiments reflect that, only utilizing the model with a small number of parameters (7B <em>v.s.</em> <span><math><mo>&gt;</mo></math></span>100B), Meta-PKE outperforms the state-of-the-art methods by near 15%, 20%, and 10% on 3 datasets, which cover not only daily but also non-daily scenarios more efficiently.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104097"},"PeriodicalIF":7.4,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DCCMA-Net: Disentanglement-based cross-modal clues mining and aggregation network for explainable multimodal fake news detection","authors":"Siqi Wei,&nbsp;Zheng Wang,&nbsp;Meiling Li,&nbsp;Xuanning Liu,&nbsp;Bin Wu","doi":"10.1016/j.ipm.2025.104089","DOIUrl":"10.1016/j.ipm.2025.104089","url":null,"abstract":"<div><div>Multimodal fake news detection is significant in safeguarding social security. Compared with single-text news, multimodal news data contains rich cross-modal clues that can improve the detection effectiveness: modality-common semantic enhancement, modality-specific semantic complementation, and modality-specific semantic inconsistency. However, most existing studies ignore the disentanglement of modality-specific and modality-common semantics but treat them as an entangled whole. Consequently, these studies can only implicitly explore the interactions between modalities, resulting in a lack of explainability. To address that, we propose a Disentanglement-based Cross-modal Clues Mining and Aggregation Network for explainable fake news detection, called DCCMA-Net. Specifically, DCCMA-Net decomposes each modality into two distinct representations: a modality-common representation that captures shared semantics across modalities, and a modality-specific representation that captures unique semantics within each modality. Then, leveraging these disentangled representations, DCCMA-Net explicitly and comprehensively mines three cross-modal clues: modality-common semantic enhancement, modality-specific semantic complementation, and modality-specific semantic inconsistency. Since not all clues play an equal role in the decision-making process, DCCMA-Net proposes an adaptive attention aggregation module to assign contribution weights to different clues. Finally, DCCMA-Net aggregates these clues based on their contribution weights to obtain highly discriminative news representations for detection, and highlights the most contributive clues as explanations for the detection results. Extensive experiments demonstrate that DCCMA-Net outperforms existing methods, achieving detection accuracy improvements of 2.53%, 4.01%, and 3.99% on Weibo, PHEME, and Gossipcop datasets, respectively. Moreover, the explainability accuracy of DCCMA-Net exceeds that of current state-of-the-art methods on the Weibo dataset.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104089"},"PeriodicalIF":7.4,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MHGC: Multi-scale hard sample mining for contrastive deep graph clustering","authors":"Tao Ren ,&nbsp;Haodong Zhang ,&nbsp;Yifan Wang ,&nbsp;Wei Ju ,&nbsp;Chengwu Liu ,&nbsp;Fanchun Meng ,&nbsp;Siyu Yi ,&nbsp;Xiao Luo","doi":"10.1016/j.ipm.2025.104084","DOIUrl":"10.1016/j.ipm.2025.104084","url":null,"abstract":"<div><div>Contrastive graph clustering holds significant importance for numerous real-world applications and yields encouraging performance. However, current efforts often overlook hierarchical high-order semantic information and treat all contrastive pairs equally during optimization. Consequently, the abundance of well sample pairs overwhelms the critical structural context learning process, limiting the accumulation of information and deteriorating the network’s learning capability. To address this concern, a novel contrastive deep graph clustering method termed MHGC is proposed by conducting hard sample mining in contrastive learning with multi-granularity. Specifically, random walk with restart is utilized to sample subgraphs centered around anchor nodes. Then, an attribute encoder to learn node representations is designed to obtain subgraph embeddings. Subsequently, hard and easy sample pairs within high-confidence clusters is identified by applying a two-component beta mixture model to the clustering loss. Building upon this, a weight regulator is then elaborated to adaptively tune the weights of sample pairs and a multi-scale contrastive loss framework is proposed to leverage structural context information in a hierarchical contrastive manner. Comprehensive experiments conducted on six widely used datasets confirm the comparable performance of our MHGC relative to the state-of-the-art baselines, demonstrating an average increase of 1.54% in accuracy. Additionally, the ablation study further proves that our proposed multi-scale learning scheme and BMM-based hard mining strategy are effective approaches for the graph clustering task. The source code is available at <span><span>https://github.com/sodarin/MHGC</span><svg><path></path></svg></span></div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104084"},"PeriodicalIF":7.4,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143403724","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Information bottleneck-driven prompt on graphs for unifying downstream few-shot classification tasks","authors":"Xin Zhang ,&nbsp;Wanyu Chen ,&nbsp;Fei Cai ,&nbsp;Jianming Zheng ,&nbsp;Zhiqiang Pan ,&nbsp;Yupu Guo ,&nbsp;Honghui Chen","doi":"10.1016/j.ipm.2025.104092","DOIUrl":"10.1016/j.ipm.2025.104092","url":null,"abstract":"<div><div>Inspired by the success of prompt in natural language processing, the graph prompt-based methods are proposed to solve the classification tasks under the conditions with limited instances. Recent graph prompt-based methods typically employ link prediction as the pre-training task, which brings a gap between the pre-training task and the downstream classification task, thus introducing irrelevant and noisy features to the downstream tasks. To tackle this issue, we propose a framework called Diffused Graph Prompt (Di-Graph), which consists of three major components, <em>i.e</em>., the diffused pre-training, a graph prompt layer, and an information bottleneck optimizer. Specifically, the diffused pre-training aims to obtain the stable node features with a diffusion process, mitigating the gap between the pre-training tasks and the downstream tasks. The graph prompt layer enhances the pre-trained model to leverage its knowledge via capturing both the structural and node features in graphs. The information bottleneck optimizer helps the model discard redundant features by retaining the minimal sufficient statistic of the input data. Extensive experimental results on five public graph datasets demonstrate that our Di-Graph model surpasses the state-of-the-art model in terms of accuracy for both node and graph classification tasks. In particular, for graph-level tasks, Di-Graph achieves an average performance gain of 6.50% over the previous best model (GraphCL) on BZR dataset. For node-level tasks, Di-Graph achieves a 2.05% improvement over the best baseline (GraphCL) on PROTEINS dataset.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104092"},"PeriodicalIF":7.4,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395729","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring the subject heterogeneity of scientific research projects funding-example of the Chinese natural science foundation","authors":"FeiFei Wang ,&nbsp;WenHua Guo ,&nbsp;Rui Xue ,&nbsp;Claude Baron ,&nbsp;ChenRan Jia","doi":"10.1016/j.ipm.2025.104098","DOIUrl":"10.1016/j.ipm.2025.104098","url":null,"abstract":"<div><div>In the increasingly competitive landscape of science and technology funding, understanding the heterogeneous factors and outcomes in funding allocation is crucial. This study proposes a research framework of subject heterogeneity to explore how the individual and combined characteristics of scholars and research topics impact funding acquisition, intensity, and performance. We use the case of the National Natural Science Foundation of China (NSFC) funding for artificial intelligence projects from 2009 to 2018 to empirically validate our framework. The findings reveal that scholars affiliated with high-level institutions, who focus on specialized areas and produce high-quality representative work, are more likely to secure funding. Unexpectedly, funding incentives did not significantly alter scholars' enthusiasm for pursuing popular topics. Moreover, the results indicate that funding has a more substantial impact on cultivating scholars than on advancing new research topics, particularly in the short term. Expanding the scope of funding proves to be more effective in enhancing research performance than merely increasing funding intensity. These insights provide valuable guidance for researchers in topic selection and submission strategies, as well as for policymakers aiming to optimize the management of competitive scientific projects.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104098"},"PeriodicalIF":7.4,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395728","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Competition or coexistence: Diffusion network differences between entertainment events and public events on social media","authors":"Sini Su ,&nbsp;Yusong Dai ,&nbsp;Xiaoke Xu ,&nbsp;Zhijin Zhong","doi":"10.1016/j.ipm.2025.104087","DOIUrl":"10.1016/j.ipm.2025.104087","url":null,"abstract":"<div><div>There is a prevalent concern that public information will be marginalized due to the prevailing preference for entertainment content on social media, consequently impacting public engagement. Despite extensive discussions, the relationship between entertainment and public affairs remains ambiguous. Unlike the majority of relative studies that examine broad phenomena or topics, we focus on event-specific diffusion networks, thereby avoiding ambiguous information categorization. Specifically, we separately selected 10 of the most influential events that happened from 9:00 to 15:00 on June 23, 2021, in both entertainment and public fields on Weibo. The collected dataset comprises 4,361,793 original posts, 16,511,446 reposts, and 10,557,370 users. Based on the diffusion network of entertainment events with those of public events, we observed that entertainment events do not divert attention from public events closely associated with people's lives. This remains the case despite entertainment events steadily exhibiting higher diffusion characteristics than those of most public events. Notably, public events can sustain public attention and discussions for longer. In addition, there are differences between the early followers of public events and entertainment events. The former predominantly comprises unverified users, while the latter mainly consists of verified users. Overall, the flow of sentiment tends to be consistent, stable, and transferable in both types of events. This study also utilized event data from June to December 2020, which underwent a complete diffusion process, to reaffirm these findings, thereby validating their explanatory power on a larger scale.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104087"},"PeriodicalIF":7.4,"publicationDate":"2025-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143387246","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"LLM-infused bi-level semantic enhancement for corporate credit risk prediction","authors":"Sichong Lu ,&nbsp;Yi Su ,&nbsp;Xiaoming Zhang ,&nbsp;Jiahui Chai ,&nbsp;Lean Yu","doi":"10.1016/j.ipm.2025.104091","DOIUrl":"10.1016/j.ipm.2025.104091","url":null,"abstract":"<div><div>Corporate credit risk (CCR) prediction enables investors, governments, and companies to make informed financial decisions. Existing research primarily focuses solely on the tabular feature values, yet it often overlooks the rich inherent semantic information. In this paper, a novel bi-level semantic enhancement framework for CCR prediction is proposed. Firstly, at the data-level, a large language model (LLM) generates detailed textual descriptions of companies’ financial conditions, infusing raw tabular training data with semantic information and domain knowledge. Secondly, to enable semantic perception during inference when only tabular data is available, a contrastive multimodal multitask learning model (CMML) is proposed at the model level. CMML leverages the semantically enhanced data from the previous level to acquire semantic perception capabilities during the training phase, requiring only tabular data during prediction. It aligns the representations of tabular data with textual data, enabling extracting semantically rich features from tabular data. Furthermore, a semantic alignment classifier and an MLP classifier are integrated into a weighted ensemble learner within a multitask learning architecture to enhance robustness. Empirical verification on two datasets demonstrates that CMML surpasses benchmark models in key metrics, particularly in scenarios with limited samples and high proportions of unseen corporations, implying its effectiveness in CCR prediction through bi-level semantic enhancement.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104091"},"PeriodicalIF":7.4,"publicationDate":"2025-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143379293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Concept cognition over knowledge graphs: A perspective from mining multi-granularity attribute characteristics of concepts","authors":"Xin Hu ,&nbsp;Denan Huang ,&nbsp;Jiangli Duan ,&nbsp;Pingping Wu ,&nbsp;Sulan Zhang ,&nbsp;Wenqin Li","doi":"10.1016/j.ipm.2025.104095","DOIUrl":"10.1016/j.ipm.2025.104095","url":null,"abstract":"<div><div>Humans can better understand and answer questions than machines because they know the cognitive knowledge related to the concept in questions. To equip machines with the cognitive knowledge required for cognizing concepts, concept cognition over knowledge graphs in this study involves mining the cognitive knowledge required by machines, i.e., multi-granularity attribute characteristics of concepts, which enables machines to distinguish or cognize concepts from multiple granularities. First, an algorithm is proposed to mine multi-granularity attributes characteristics of concepts from concept-related knowledge in a knowledge graph, i.e., frequent attributes and attribute values of concepts from multiple granularities. Second, the monotonicity of the multi-granularity attribute pattern is proposed to promote synergy among granularities and accelerate the mining process because the result from coarser granularity can serve as a candidate for the result from finer granularity. Third, the representativeness of the maximal frequent attribute pattern is used to unleash the value of above monotonicity and accelerate the mining process, which enables the algorithm to mine maximal frequent attribute patterns with fewer quantities to derive all frequent attribute patterns in large numbers. Finally, the experiments show that the above algorithm is more efficient than baseline algorithms, the monotonicity of the multi-granularity attribute patterns can accelerate the mining process, the representativeness of the maximal frequent attribute patterns means that the percentage is always less than 5%, the percentages of correctly classified instances by the multi-granularity attribute characteristics are always higher than 90%, and the above classification performance performs better than existing machine learning algorithms at most cases.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104095"},"PeriodicalIF":7.4,"publicationDate":"2025-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143379295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An LLM-assisted ETL pipeline to build a high-quality knowledge graph of the Italian legislation","authors":"Andrea Colombo,&nbsp;Anna Bernasconi,&nbsp;Stefano Ceri","doi":"10.1016/j.ipm.2025.104082","DOIUrl":"10.1016/j.ipm.2025.104082","url":null,"abstract":"<div><div>The increasing complexity of legislative systems, characterized by an ever-growing number of laws and their interdependencies, has highlighted the utility of Knowledge Graphs (KGs) as an effective data model for organizing such information, compared to traditional methods, often based on relational models, which struggle to efficiently represent interlinked data, such as references within laws, hindering efficient knowledge discovery.</div><div>A paradigm shift in modeling legislative data is already ongoing with the adoption of common international standards, predominantly XML-based, such as Akoma Ntoso (AKN) and the Legal Knowledge Interchange Format, which aim to capture fundamental aspects of laws shared across different legislations and simplify the task of creating Knowledge Graphs through the use of XML tags and identifiers. However, to enable advanced analysis and data discovery within these KGs, it is necessary to carefully check, complement, and enrich KG nodes and edges with properties, either metadata or additional derived knowledge, that enhance the quality and utility of the model, for instance, by leveraging the capabilities of state-of-the-art Large Language Models.</div><div>In this paper, we present an ETL pipeline for modeling and querying the Italian legislation in a Knowledge Graph, by adopting the property graph model and the AKN standard implemented in the Italian system. The property graph model offers a good compromise between knowledge representation and the possibility of performing graph analytics, which we consider essential for enabling advanced pattern detection. Then, we enhance the KG with valuable properties by employing carefully fine-tuned open-source LLMs, i.e., BERT and Mistral-7B models, which enrich and augment the quality of the KG, allowing in-depth analysis of legislative data.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104082"},"PeriodicalIF":7.4,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Improving cross-document event coreference resolution by discourse coherence and structure","authors":"Xinyu Chen,&nbsp;Peifeng Li,&nbsp;Qiaoming Zhu","doi":"10.1016/j.ipm.2025.104085","DOIUrl":"10.1016/j.ipm.2025.104085","url":null,"abstract":"<div><div>Cross-Document Event Coreference Resolution (CD-ECR) is to identify and cluster together event mentions that occur across multiple documents. Existing methods exhibit two limitations: (1) In contrast to within-document event mentions, which are linked by rich, coherent contexts, cross-document event mentions lack such contexts, posing a challenging for the model to understand the relation between two event mentions in different documents. (2) The lack of coherent textual information between cross-document event mentions lead to the inability to capture their global information, which is important to mine long-distance interactions between them. To tackle these issues, we propose a novel discourse coherence enhancement mechanism and introduce discourse structure to improve cross-document event coreference resolution. Specifically, we first introduce a new task: Event-oriented cross-document coherence enhancement (ECD-CoE), which selects coherent sentences that form a coherent text for two cross-document event mentions. Second, we represent the coherent text as a tree structure with rhetorical relation information between textual units. We then obtain the global interaction information of event mentions from the tree structures and finally resolve coreferent events. Experimental results on both the ECB+ and GVC datasets indicate that our proposed method outperforms several state-of-the-art baselines.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"62 4","pages":"Article 104085"},"PeriodicalIF":7.4,"publicationDate":"2025-02-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}