RADAR Release v1.26: Improved Data Upload, AI-Powered Keyword Expansion, and New User Role

As of November 05, 2026, RADAR Cloud is available in software version v1.26.2 (frontend) / v1.19.0 (backend).

With this new software release, we have comprehensively expanded RADAR’s functionality in several areas. The highlights are a fundamentally redesigned data upload feature, which significantly simplifies the transfer of files and entire folder structures, and an optional AI-powered keyword expansion feature that supports improved metadata annotation.

With the new “Dataset Curator” user role and enhanced access rights management, we also enable more flexible collaboration on individual datasets. Furthermore, we are improving the interoperability and discoverability of research data by supporting the additional metadata formats xMetaDissPlus and DCAT, as well as the new “Contributors” filter facet in the publication list.

The new features in detail:

With the new software release, we have completely redesigned the data upload process via the web interface, significantly simplified it, and optimized it for reliable data transfer. Whereas previously only individual files could be uploaded and entire folders had to be packed into archive files (e.g. .zip) first, the new upload form offers much more flexible and intuitive options.

Starting now, multiple files and entire folders can be uploaded simultaneously - either individually or in any combination. Selection is done as usual via drag-and-drop or through the file manager. The original folder structure is preserved, so even large datasets can be imported directly into RADAR without first being packed into archive files. These flexible upload options are also available for datasets that have already been created and are in the “Pending” status, allowing you to add additional files and folders.

The user interface has also been extensively modernized. Selected files and folders are displayed in a clear, hierarchically organized list before upload. Individual folders can be expanded or collapsed to maintain an overview even with comprehensive data sets. Files and folders can be removed from the selection as needed. In addition, the upload form displays the number of files and folders to be uploaded, as well as their total data volume.

For transfers via the browser, RADAR Cloud now limits uploads to a maximum of 1,000 files and 25 GB per upload operation. Different limits may apply for RADAR Local. The currently applicable upload limits are displayed directly in the form. If the available storage quota is less than the respective upload limit, the lower value is displayed. The upload limits help ensure reliable data transfer via the browser and make it easier to choose the appropriate upload method. For larger data volumes and more extensive file collections, WebDAV and die RADAR REST API are still available.

RADAR now offers an optional AI-assisted feature for expanding keywords in the metadata editor. The feature can be accessed via the “Extend Keywords with AI (Beta)” button in the right-hand navigation frame. The new RADAR Keyword Service makes it possible to identify additional relevant keywords based on existing (meta)data. In this way, RADAR supports the subject-specific indexing of research data and improves its discoverability.

The RADAR Keyword Service is based on the KeyBERT analysis method, which uses BERT language models (Bidirectional Encoder Representations from Transformers) to extract relevant keywords and word groups from texts. The BERT language models take linguistic context into account, while KeyBERT evaluates the terms based on their content-related relevance. Generative AI is not used in this process. Processing takes place on an AI infrastructure hosted by FIZ Karlsruhe itself.

For keyword analysis, there are currently just under 20 different BERT language models available, each with a different subject focus. Users can select a suitable model, specify the maximum number of keywords to be identified, and decide whether a README file should be included in the analysis in addition to the metadata (title and description). Based on this, the RADAR Keyword Service extracts relevant terms and word groups. These are then matched against standardized terminologies from an ontology collection previously selected by the user via the TS4NFDI Terminology Service to suggest appropriate standardized keywords. The suggested keywords are highlighted in color in the metadata editor. Users can review them, remove them if necessary, and save them to incorporate them into the metadata. The decision on whether to add the keywords is thus entirely up to the users ("human-in-the-loop").

The new optional feature is initially being released as a beta version and is disabled by default. Contract administrators can enable it selectively for individual or all workspaces. In the future, the selection of available language models may be expanded, also in consultation with scientific communities.

With this new software release, RADAR is introducing the “Dataset Curator” user role. At the same time, the existing option to share datasets with other users is being expanded and consolidated into a centralized access rights management system.

Administrators and curators can now grant specific permissions for individual datasets and determine who is allowed to edit them or access them in read-only mode. Two roles are available for this purpose:

• Dataset Readers can access a dataset in read-only mode but cannot make any changes. This option was previously available for archived datasets as well as for published datasets under embargo. It is now also being extended to datasets with the status “Pending”.
• Dataset Curators can edit a dataset and perform the available actions, such as editing metadata, uploading research data, initiating a peer review, archiving or publishing the dataset, and correcting metadata retroactively. Unlike curators of a workspace, their permissions are limited to the specific dataset. Access to other datasets within the same workspace is therefore not granted.

Permissions are managed directly within the dataset: In the right-hand navigation column, under “Access and Permissions,” you’ll find the “Manage Access Rights” button. Using the corresponding form, registered RADAR users can be assigned either of the two roles, and permissions can be revoked at any time.

Datasets for which users have been granted the appropriate permissions are listed on the dashboard under “Datasets with Direct Access" with the respective assigned role after login. They can be accessed directly from there without requiring access to the associated workspace.

The new “Data Set Curator” role thus enables targeted collaboration on individual datasets without having to grant access to the entire workspace. This offers additional flexibility, particularly for cross-institutional research projects or API-based workflows.

RADAR now supports two additional metadata formats: xMetaDissPlus and DCAT. This brings the total number of metadata export options available on the landing pages for published data sets to twelve. In the future, both formats will also be offered via the RADAR OAI provider for automated metadata retrieval (harvesting).

• xMetaDissPlus is an XML-based exchange format developed by the German National Library (DNB) in collaboration with the Baden-Württemberg Library Service Center (BSZ). It is designed for the standardized exchange of bibliographic metadata, including metadata on research data, and can be used, for example, to import metadata into research information systems such as HISinOne-RES. 
   
• DCAT (Data Catalog Vocabulary) is an RDF vocabulary standardized by the World Wide Web Consortium (W3C) for describing datasets in data catalogs. It facilitates the exchange and integration of metadata from different sources and improves the discoverability of datasets across various portals, for example in the open data sector.

The enhancement was initiated by the University of Würzburg, which operates a local RADAR instance called WueData and also developed the initial metadata transformation using XSLT stylesheets.

The RADAR publication list can now also be filtered by individuals and institutions that have contributed to a dataset as “contributors.” This new facet complements the existing filters “Creator,” “Publisher,” and “Rights Holder.” This brings the total number of filter options available for narrowing down the published datasets to twelve.
 

Note: This software release also affects the community-specific publicaion services for research data based on RADAR:

• RADAR4Chem, funded by NFDI4Chem (DFG Project Number 441958208)
• RADAR4Culture, funded by NFDI4Culture (DFG Project Number 441958017)
• RADAR4Memory, funded by NFDI4Memory (DFG Project Number 501609550)