Skip to main content

Release Notes - PoolParty 2024 R1 (9.3.0-9.3.1)

03/06/2024

Our latest PoolParty release enhances user experience by introducing the new beta feature Taxonomy Advisor, which helps subject matter experts and taxonomists add relevant narrower concepts more efficiently using Large Language Models (LLMs). This release also includes the Inference Tagging feature - a rule-based expansion of the extracted concepts - paired with the Visual Query Builder feature to assist users less familiar with SPARQL queries; this all being complemented by numerous improvements and security enhancements.

Highlights

PoolParty's beta feature, Taxonomy Advisor, enhances taxonomy enrichment by leveraging Large Language Models (LLMs) to create narrower concepts and alternative labels. By employing the LLM and carefully crafted prompts, the solution generates relevant concept candidates and alternative labels while upholding stringent security standards. The upcoming fall release will introduce additional capabilities, including bulk processing and corpus operations.

Inference Tagging goes beyond explicit concepts by deriving new information from implicit concepts in documents using existing datasets and specified rules. It leverages ontologies and rules to reveal hidden relationships, enabling interpretation, conclusions and predictions. This process is vital for automated decision-making, helping computing effectively understand complex data patterns and relationships.

The Recommender Workbench has now evolved into a new Workbench providing tools to users of Inference Tagging as well as Recommendation services (depending on the license in place). Its overhauled UI can be used to test, review and store extraction, expansion and recommendation settings. The Visual Query Builder is available to subject matter experts to assist them in writing expansion SPARQL queries.

Improvements

Improvements implemented in the current release include:

  • The project creation dialog and workflow have been revamped to streamline them onto a single screen, guiding users through all necessary information. The configuration summary report is shown at the end for users to review before finalizing the project creation.

  • We enhanced the password generation policies for new user accounts by creating longer and more complex passwords. This update ensures that the user creation process functions smoothly even when password policies are adjusted for increased security in the future (Configuring Password Policies (Keycloak)).

  • We are always focused on improving performance and have made significant enhancements in this release. The concept extraction performance has more than doubled, and the thesaurus tree performance has been improved for better user experience with larger trees.

  • The project creation API now enables users to specify the project UUID identifier for the new project. This enhancement facilitates seamless access to multiple instances of the same project in replication scenarios.

  • The multi-select frontend element design has been enhanced to support a larger number of elements.

  • The new Workbench now also supports use of files and URLs as input for concept extraction, expansion and recommendation. The Workbench functionality is linked to the PoolParty license variant - ranging from basic user interface for concept extraction through Inference Tagging all the way down to Recommendation.

Security Fixes

The following vulnerabilities have been successfully addressed:

Bug Fixes

  • User management now supports display of more than 100 PoolParty users.

  • The URIs of concept schemes can now be edited.

  • The visual browser of the Linked Data frontend works as intended when a custom scheme is enabled in the project.

  • Previously, concept extraction across multiple projects was sometimes unreliable due to extraction scores being influenced by the order of project IDs, this is now working as intended.

  • Extraction now works as intended with PPTX files.

  • The application of project user roles now functions as intended.

  • Fixed the issue where importing SKOS matching relations (such as skos:exactMatch or skos:narrowerMatch) using a tabular import resulted in empty lines being displayed in the project detail view.

  • The SPARQL update endpoint has been fixed to support updates using form-data and URL-encoded formats, which were previously failing.

  • In GraphSearch, facet values were being incorrectly formatted with added language tags and data types, potentially disrupting integrations and UI display. This issue has been resolved by returning only the actual values, with data types now provided in a separate field.

  • Corpora now work as intended in languages with regional variants (locales), such as British English (en-GB) and American English (en-US).

Component Versions

  • Tomcat: 9.0.87

  • Keycloak: 22.0.5

  • RDF4J: 4.3.6

  • Spring Core: 5.3.32

  • Spring Security: 5.8.10

  • Tika: 2.9.1

  • Elasticsearch: 8.12.1

  • Solr: 9.4.0

Note

Please note that the GraphSearch content create API will be deprecated with the next release. This will make the 'facets' parameter work differently.

The Keycloak Mapper type 'SAML Attribute to Group' will be deprecated with the next release. Instead the Keycloak proprietary Mapper type 'Advanced Attribute to Group' will be used. We therefore recommend using this mapper type from now on.

Note

Starting with PoolParty 2024 Release 1 (9.3.1) the on-the-fly migration is no longer supported since it is assumed that all users have already been migrated from PoolParty version 8.x to at least the PoolParty 2022 Release 1 (9.0.) version. It is therefore essential to make sure that all users have been migrated from version 8.x to PoolParty 2022 Release 3 (9.2.1) or lower before upgrading to version PoolParty 2024 Release 1 (9.3.1).

In this section: