Source-linked AI summary
Westlake Scholar: AI-Enhanced Scholarly Discovery over an Institutional Repository
Junshu Pan, Luodan Zhang, Yifeng Lu, Mengfan Zhao, Ming Luo, Zijie Yang, Yue Zhang, Rui Shang
TL;DR
Institutional repositories preserve and disseminate research outputs but offer limited support for connecting deposited papers to related research and people. Westlake Scholar adds four AI services over a shared, institution-grounded knowledge layer connecting approved records, paper content, and scholar–publication relationships. Its deployment at Westlake University since April 2026 demonstrates that the integrated system can operate in a live institutional setting while retaining provenance, human review, and institutional governance.
Problem
Institutional repositories provide mature preservation and dissemination infrastructure, but conventional interfaces offer limited support for connecting deposited papers to related research and people.
Method
Westlake Scholar combines repository infrastructure with contextual reading, paper discovery, publication-grounded expert discovery, and AI-generated research chronologies over a shared institutional knowledge layer.
Results
The Westlake University deployment has operated since April 2026, demonstrating that the shared data foundation and four interfaces can operate together in a live institutional environment.
Takeaways & Limitations
The platform extends the institutional repository from deposit and access toward interactive scholarly discovery through an open-source, inspectable, and adaptable institution-controlled architecture.
Takeaways & Limitations
Until scholar-facing consent, correction, and opt-out controls are added, institutions should provide notification and correction, suppression, and appeal channels and exclude generated outputs from personnel evaluation.
Abstract
from arXiv · showhide
Institutional repositories (IRs) provide mature infrastructure for preserving and disseminating research outputs, but conventional record- and document-centric interfaces provide limited support for connecting deposited papers to related research and people. We present Westlake Scholar, an open-source, institution-grounded platform that adds four complementary artificial intelligence (AI) services to repository infrastructure: contextual paper reading, research-direction-guided paper discovery, publication-grounded expert discovery, and AI-generated research chronologies for scholars. The services draw on a shared institutional knowledge layer connecting approved publication records, paper content, and scholar--publication relationships. This allows the same paper to support contextual reading, cross-paper discovery, expert matching, and longitudinal views of scholarly work. Westlake Scholar provides an open and governable implementation of an institution-controlled AI layer that connects repository content, scholarly discovery, and researcher relationships while preserving provenance, human review, and institutional governance. A deployment at Westlake University, in operation since April 2026, demonstrates that the integrated system can operate in a live institutional setting.
1 Introduction
Westlake Scholar addresses the post-deposit gap in institutional repositories by adding four AI-enhanced scholarly services over a shared institutional foundation. The platform combines repository functions with institution-controlled discovery while preserving source context, organizational scope, provenance, review, and governance.
- Motivation: Institutional repositories preserve and disseminate research outputs, but record- and document-centric interfaces provide limited support for discovering and using related research and people.The post-deposit gap includes interpreting unfamiliar papers, tracing collection-wide relationships, locating expertise, and understanding scholars’ evolving interests.
- Motivation: The design question is how an institution-controlled data layer can support related scholarly tasks while preserving source context, organizational scope, and content governance.Retrieval-augmented language models and scholarly graphs provide relevant technical foundations for this design.
- Contribution: Westlake Scholar combines repository depositing, cataloging, searching, browsing, and access with four AI-enhanced services over a shared institutional foundation.The services support contextual paper reading, cross-paper discovery, publication-grounded expert matching, and longitudinal interpretation.
- Contribution: The shared foundation connects approved publication records, paper content, and scholar–publication relationships for all four services.This common layer supports movement from a single paper to related work, institutional expertise, and research trajectories.
- Contribution: Institutions control content approval, model processing, and chronology publication, while provenance and human review support inspection and correction.Governance is applied across the service lifecycle rather than limited to a single interface.
2 Related Work
Related work frames repositories as evolving research infrastructure and describes AI-supported, source-linked discovery, while Westlake Scholar unifies these capabilities in an institution-operated architecture. Its system-level contribution is a shared grounding layer spanning retrieval, semantic discovery, expert recommendations, and research chronologies.
- Institutional repositories: Research on institutional repositories emphasizes participation, service design, impact, user-centered infrastructure, and discoverability beyond repository interfaces.Prior work identifies awareness, attitudes, web visibility, and external indexing as relevant concerns.
- Research information management: Research information management systems aggregate publications, profiles, projects, reporting, and workflows, while VIVO represents researchers, scholarly works, expertise, and organizations.These systems extend institutional research information beyond repository deposit and access.
- AI-supported discovery: Library-native AI discovery combines natural-language inquiry with source-linked synthesis, including search translation, reranking, and answers synthesized from selected abstracts.Implementation studies also raise concerns about metadata and recommendation accuracy.
- AI-supported discovery: OpenScholar and STORM illustrate retrieval-backed, citation-supported, multi-perspective, and structured approaches to knowledge curation.A library and information science perspective links AI support to stages and uncertainties in information seeking.
- System-level contribution: Westlake Scholar brings retrieval-augmented interaction, semantic discovery, expert recommendations, and research chronologies into an institution-operated architecture.Approved repository records and scholar links provide shared grounding across all four tasks.
3 System Design and Implementation
Westlake Scholar implements four task-routed AI workflows over institution-controlled repository data, using shared records and relationships with task-specific prompts and retrieval behavior. The design combines familiar repository interaction with paper-content indexing, hybrid retrieval, expert discovery, and chronology generation.
- Architecture: The platform is a web application and service API backed by institution-controlled storage and hybrid lexical–vector retrieval, with deployment scoped to one institution.Westlake University Library, Airalogy, and the university AI laboratory co-developed the system with distinct service, engineering, and AI roles.
- Architecture: Four task handlers support paper-context chat, paper retrieval, scholar recommendation, and chronology generation while sharing records and relationships.Task-specific prompts and candidate limits preserve workflow-specific behavior within a common service layer.
- Interface design: The interface retains persistent navigation, search and browse entry points, record views, and an inline document reader while adding conversational actions.This follows an interface-consistency principle intended to align the system with users’ existing expectations.
- Retrieval and model configuration: Indexing extracts approved PDF text and combines it with titles and abstracts to create an institution-grounded evidence base.The indexed text is divided into 1,000-character segments with 100-character overlap, and segments store normalized term frequencies for BM25 ranking.
- Retrieval and model configuration: Full-text content remains within the institutional deployment by default, while only titles and abstracts are sent for embedding unless administrators authorize approved full-text processing.The system supports compatible chat and embedding models and allows a separate chronology model.
- Paper discovery: Paper recommendation fuses BM25 and vector rankings over approved institutional paper text and returns passages and identifiers from the top three papers as evidence.The language model then explains the relevance of candidate papers.
- Expert discovery: Expert discovery retrieves semantically related publication–scholar records from a research direction, paper title, abstract, or question, incorporating recent user messages into the query.Recommendations are restricted to scholars supported by retrieved records.
- Research chronologies: Chronology generation groups linked publications into fixed five-year calendar windows, checks DOI metadata against OpenAlex and Crossref, and produces focus summaries and tags from titles and abstracts.Unresolved years and conflicts are recorded for review.
4 Initial Institutional Deployment
Westlake Scholar has operated at Westlake University since April 2026, where its shared data foundation and four interfaces operate together in a live institutional environment. The deployment also connects technical integration to participatory service processes involving librarians and university users.
- Deployment: Since April 2026, Westlake Scholar has operated at Westlake University with its shared data foundation and four interfaces functioning together in a live institutional environment.The deployment provides an institutional setting for the integrated platform.
- Participatory service process: The deployment situates the platform within information-science concerns about awareness, onboarding, and sustained faculty participation in repository services.These concerns informed the service process surrounding the deployment.
- Participatory service process: Librarians and university users need accessible channels to identify needs, question generated content, and shape later iterations.This participatory-design commitment connects technical integration with continuing institutional interpretation, governance, and improvement.
5 Ethics, Governance, and Limitations
Westlake Scholar applies institutional review and provenance practices to AI-generated expert recommendations and research chronologies. Its ongoing deployment is intended to support evaluation of usage, retrieval quality, generated outputs, and scholars’ responses.
- Ethics and Governance: Generated chronologies and expert recommendations require provenance, correction, and institutional review because they characterize identifiable faculty members.Papers enter public search and recommendation only after institutional content review; metadata conflicts and unresolved publication years are recorded for chronologies.
- Limitations: Until scholar-facing consent, correction, and opt-out controls are added, institutions should provide notification, correction, suppression, and appeal channels.Generated summaries and expert rankings should also be excluded from personnel evaluation.
- Ongoing Evaluation: The ongoing institutional deployment provides a setting to evaluate usage, retrieval quality, generated outputs, and scholars’ responses.Findings from this evaluation are intended to inform subsequent design and governance.
6 Conclusion
Westlake Scholar uses an institution-controlled knowledge foundation to connect contextual reading, paper discovery, expert matching, and longitudinal scholarly views. Its open-source implementation preserves provenance, review, and institutional control while the ongoing deployment supports continued evaluation and participatory governance.
- Conclusion: An institution-controlled knowledge foundation connects contextual reading, paper discovery, expert matching, and longitudinal views of scholarly work.The integrated architecture combines repository content and scholar relationships with provenance, review, and institutional control.
- Conclusion: The open-source implementation makes the architecture inspectable and adaptable.
- Conclusion: The ongoing deployment provides a setting for continued evaluation and participatory governance.