• 学术搜索
  • 科研智能体
    • Research Labs
    • AI 阅读
    • AI 文库
    • 深度研究
    • 学者亮点
  • 学术资源
    • AI2000
    • 期刊/会议
    • 学者库
    • 学术API
      MCP&Skill
    • 溯源树
    • 数据集
  • 知识沉淀
    • 学术空间
订阅小程序
旧版功能
aminer vip
开通会员低至0.73元/天
一次搞定AI科研
立即登录
  • English
  • 联系方式
    International Journal on Digital Libraries

    International Journal on Digital Libraries

    JournalISSN 1432-5012eISSN 1432-1300中科院 管理学 4区

    年发文量

    研究主题

    论文(620)

    排序
    1Structuring Multimodal Cultural Data for Digital Memory Construction: a Hierarchical Semantic Organization Framework for User-Oriented Navigation
    Fan Liu,Zhong Tie, WenQiang Wang

    Multimodal cultural data can be preserved and retrieved, yet they often remain difficult to understand as continuous relational contexts. This study proposes a hierarchical semantic organization framework for digital memory construction and examines how it connects back-end data organization with front-end user navigation. An exploratory design research approach was adopted. A Unity-based interactive prototype was developed from 113 world exposition-related event records and more than 3,000 image resources. Nine adult participants completed four tasks—structure identification, relation tracing, immediate path reconstruction, and confusion and recovery—each lasting approximately 5–8 min, followed by a post-task interview of approximately 5–10 min. Contemporaneous observation notes and written post-task interview records were analyzed through directed content analysis. Five participants began with a year or temporal entry, two used a single non-temporal entry, and two switched among multiple entries. Six participants explained at least one cross-event or cross-object relational context, six demonstrated multimodal contextualization, and six completely or substantially reconstructed their browsing paths immediately after the tasks. One participant showed a clear navigation difficulty; another briefly paused under high node density but did not meet the coding threshold for navigation difficulty. The study proposes a layered navigation mechanism that transforms uneven data coverage and depth into structures that are locatable, relatable, accessible, and returnable, while clarifying design boundaries concerning relation interpretability, current-position cues, return mechanisms, and information-density control.

    2026
    引用
    AI阅读
    加入学术空间
    2Weaponizing Cultural Heritage in the Digital Age: Evidence from Wikipedia Revisions
    Hamest Tamrazyan,Maxime Garambois,Emanuela Boros

    In contemporary hybrid conflicts, cultural heritage is targeted not only through physical destruction but increasingly also through digital manipulation. This study proposes a systematic framework for detecting and analyzing cultural weaponization such as the strategic distortion, reframing, or erasure of cultural narratives, on English Wikipedia, using Ukraine as a case study and combining large-scale revision analysis, semi-automatic LLM-based annotation, and expert validation to trace how contested cultural meanings evolve over time. Our contributions include a conceptual framework that distinguishes cultural weaponization from related phenomena such as propaganda, disinformation, and censorship, positioning it as a distinct form of symbolic and epistemic intervention, as well as a structured dataset of more than 70,000 culturally and politically salient Wikipedia revisions supported by a validated gold-standard subset. We further develop an LLM-assisted analytical pipeline capable of identifying subtle manipulative strategies, including selective omissions, terminological shifts, narrative reframings, and structural edits. Our findings show that cultural weaponization on open platforms rarely relies on explicit falsehoods, but instead emerges through accumulative, procedurally legitimate edits that reshape evidentiary authority and historical interpretation. From these patterns, we derive an emerging typology of epistemic, narrative, and structural weaponization, highlighting the need for context-sensitive approaches to safeguarding digital cultural heritage.

    2026
    引用
    AI阅读
    加入学术空间
    3Schema, Values, and Priors: Understanding LLM-based Domain Detection for Tabular Datasets
    Elizaveta Gamper, Roman Netrogolov,Anna Kalyuzhnaya,Irina Deeva

    Automatic identification of the semantic domain of a tabular dataset is an important step toward dataset discovery, metadata enrichment, and domain-aware data assistants. In this paper, we study dataset-level domain detection for tabular data using large language models (LLMs), classical machine learning baselines, and hybrid prior-guided inference. We construct an OpenML-derived benchmark by transforming noisy user-provided tags into an interpretable 14-domain taxonomy through embedding-based clustering, and we complement it with a second benchmark based on native categories from the NYC Open Data portal. Using these benchmarks, we evaluate gpt-4o and DeepSeek-V3 under zero-shot, one-shot, and few-shot prompting regimes (with no examples, one or several examples), and compare them with non-LLM baselines. The results show that LLMs can infer dataset domains from compact textual representations and are competitive with specialized baselines, while column-name embedding classifiers provide strong ranked candidates and higher Top-3 recall in several settings. We further introduce a hybrid approach in which probabilities from a Random Forest classifier trained on column-name embeddings are injected into the LLM prompt as structured prior information. Finally, we analyze representation effects and feature-type sensitivity. Column headers alone often provide sufficient semantic signal, small row samples are usually as effective as larger ones, and categorical string features improve domain recognition only when they contain meaningful domain-specific cues. Overall, the study shows that reliable tabular domain detection depends not only on the model, but also on the semantic quality of schema elements, cell values, benchmark label spaces, and auxiliary classifier priors.

    2026
    引用
    AI阅读
    加入学术空间
    4Diagnosing the Limits of Large Language Models for Structured Scientific Evidence Extraction
    Zhiyin Tan,Jennifer D’Souza,Sören Auer

    Scientific evidence is predominantly communicated through unstructured articles, while many downstream workflows require structured representations of study populations, variables, methods, and quantitative findings. Large language models (LLMs) offer a promising route toward automating this transformation, yet their reliability in constructing schema-constrained evidence records from parsed scientific full texts remains insufficiently understood. Evidence synthesis is a representative use case in which systematic reviews and meta-analyses require accurate, structured study information before quantitative integration can begin. We present a schema-grounded, tiered evaluation framework for structured scientific evidence extraction that increases task difficulty from single-entity identification to relational tuples and corpus-level statistical tasks. We evaluate frontier commercial and open-weight models across five scientific domains under a stricter multi-level assessment protocol, complemented by human validation of model-based judging and process-level inspection of intermediate reasoning for statistical tasks. Results reveal a clear pattern: models perform substantially better on low-complexity extraction tasks than on tasks requiring correct bindings among entities, numerical values, and methodological attributes. Intermediate-step analysis further indicates error amplification under exact downstream aggregation, where small upstream errors often prevent correct final outputs. Scaling model size yields uneven gains and does not remove these weaknesses. These findings show that recent advances have not resolved the fundamental challenges of structured scientific evidence extraction. We provide a reusable query suite and diagnostic evaluation framework to support future fine-grained benchmarks and more rigorous assessment of models for evidence-intensive scientific workflows.

    2026
    引用
    AI阅读
    加入学术空间
    5Bridging SPARQL Endpoints and Scholarly Infrastructures: an Interoperability Suite for RDF Cultural Heritage Data
    Giorgia Rubin,Alessia Bardi, Riccardo Del Gratta,Angelo Mario Del Grosso,Miriam Baglioni, Andrea Dell’Amico

    RDF knowledge graphs and SPARQL endpoints are widely adopted in the Cultural Heritage and Digital Humanities domains to model semantically rich research metadata. However, SPARQL-based access remains poorly aligned with the interoperability requirements of large scholarly aggregation infrastructures and with developer practices centered on RESTful services, limiting the integration and reuse of RDF-native repositories. This paper presents the interoperability suite developed within the DH-ATLAS project, which produced an ontology and a knowledge graph describing Digital Humanities research on Italian Cultural Heritage. The suite includes two complementary components. A configurable OAI-PMH publisher exposes RDF metadata as records compliant with the OpenAIRE Guidelines v4, enabling the harvesting and ingestion of DH-ATLAS resources into aggregators such as the OpenAIRE Graph. In addition, a RESTful API based on the RAMOSE framework offers simplified, developer-friendly access to RDF content, reducing the need for SPARQL expertise. Beyond technical interoperability, the paper illustrates how the DH-ATLAS Knowledge Graph can support thematic aggregation at scale. We describe the knowledge-graph-driven process used to define a subcommunity on Italian Cultural heritage within the OpenAIRE CONNECT Gateway on Digital Humanities and Cultural Heritage, combining expert curation with scalable named-entity-based extraction techniques. The proposed approach enhances the accessibility, interoperability, and reuse of RDF cultural heritage data and provides a reusable model for Digital Humanities initiatives integrating with large-scale scholarly infrastructures.

    2026
    引用
    AI阅读
    加入学术空间
    立即登录,查看全部 620 篇论文

    高被引作者

    作者引用发文
    Tova Milo9972
    Bela Gipp9413
    Sophia Ananiadou8741
    Katerina T. Frantzi8741
    Serge Abiteboul7643
    Stefan Langer7191
    Corinna Breitinger7191
    Joeran Beel7191
    Alberto Mendelzon6861
    George Mihaila6861

    高产作者

    作者引用发文
    Michael Nelson28914
    Edward A. Fox39611
    Philipp Mayr1569
    Annika Hinze619
    Herbert Van De Sompel1688
    Yannis Ioannidis1258
    Michele C. Weigle1898
    Andreas Rauber1367
    Adam Jatowt1367
    Marcos André Gonçalves1936

    相关期刊

    Data Intelligence
    Journal of Information Science
    International Journal of Metadata, Semantics and Ontologies
    Theory and Practice of Digital Libraries
    Digital Humanities in the Nordic Countries
    International Conference on Asian Digital Libraries
    International Symposium of Information Science
    Italian Research Conference on Digital Library Management Systems
    Journal of Data and Information Science
    ACM/IEEE Joint Conference on Digital Libraries