Graph Modeling and Semantic Analytics 10.2 Release Notes

The user interface of the PoolParty Suite is undergoing a comprehensive transformation to align with its new corporate identity: the Graphwise Platform. This transformation represents a unified evolution toward a more sophisticated and intuitive user experience. In the initial phase of this transition, the core sub-components of PoolParty have been rebranded as Graph Modeling. This strategic advancement embodies the broader corporate vision of Graphwise to deliver comprehensive, enterprise-grade AI solutions underpinned by a robust semantic foundation.

PoolParty 10.2 introduces a Lucene-based tagging engine delivering native multilingual document processing across over 40 languages. The solution employs Lucene language analyzers and Finite State Transducers for sophisticated tokenization and inflection handling, eliminating dependencies on external language services or translation-based approaches.

To activate, set the extractorType parameter to "LUCENE" in API requests. Taxonomies must contain labels in target languages. This engine will be the default in future releases.

Key Capabilities

Native Japanese & Multilingual Support: Native processing for Japanese, Russian, Chinese, and other languages via Lucene language analyzers.
Unified API (/api/tag): Consolidates extract, annotate, categorize, and zip-extract functions into a single high-performance endpoint. Discontinued legacy NIF support.
Multi-Document Processing: In aggregate mode, zip container contents are treated as unified text. This supports multiple direct file inputs per request.
Automatic Indexing: After taxonomy changes, extraction models (FST) are automatically rebuilt, eliminating manual refresh requirements.
Reduced Architectural Dependency: The tagging engine no longer relies on external services for its lookup logic by leveraging Lucene FSTs.

Known Differences

Scoring Variance: Lucene's punctuation and term-weighting treatment differs from the classic engine, producing score variations that reflect architectural differences rather than any decline in quality.
Case Sensitivity: Enhanced acronym handling maintains case sensitivity for uppercase labels while improving multi-word concept matching in capitalized text.

GraphSearch Exact Phrase Matching enhances search precision by ensuring exact matches rank first. This capability supports specialized research workflows that require isolation of specific technical terms.

The new API-level configuration enables users to activate Exact Phrase Matching for designated search filters, ensuring the search engine identifies contiguous token sequences in the precise specified order.

Key Capabilities

Contiguous Token Matching: Leverages Elasticsearch match_phrase with slop: 0 to enforce strict token adjacency.

Mandatory vs. Optional Modes

Mandatory Mode: Returns empty results when no exact phrase match is found.
Optional Mode: Applies significant ranking boosts to exact matches while maintaining visibility of other relevant content.

Specialized Term Support: Provides enhanced handling of alphanumeric strings that standard analyzers would previously tokenize inappropriately.

Known Limitations

This feature operates within existing analyzer constraints, resulting in case-insensitive matching that disregards delimiters such as hyphens, colons, or superfluous spaces.

High-frequency stop words (such as "the" or "and") may be excluded during phrase matching, potentially affecting accuracy.

We have introduced OAuth 2.0 client credentials support for communication between Graph Modeling and GraphDB. This update allows Graph Modeling to authenticate securely using Bearer tokens obtained from an authorization server such as Keycloak moving away from legacy HTTP Basic Authentication.

The implementation includes automatic token retrieval, caching, and background refresh to ensure seamless, secure data access while maintaining backward compatibility with existing authentication methods.

Document content now displays correctly when using a trained Classifier.
Terminology Import Assistant now works as intended.
InScheme concept colors now update correctly.
Fixed duplicate projects appearing on the Open Project page.
Collections now open without triggering any server errors in special scenarios..
System requirement warnings in Corpus Management now display correctly.
The full set of special characters is now supported by Webhook notifications.
The Search feature now provides relevant results after a user has entered a keyword in the search bar also for terms not found using the autocomplete functionality.
An error returned for some concept schemes when clicking it on the UI resulting in the user being thrown out from the project has been remedied.
The expansion function in the Taxonomy Builder for existing concepts now works as intended.
The bug preventing SSO idle timeout from Keycloak configurations has been resolved.
Adding a document to a training box no longer returns a 405 error (method not allowed).
An issue triggering an error when searching for concept on the Advanced Search tab has been resolved.
Piwik Pro Analytics has been added to the CSP policies addressing CSP specific restrictions and now works as intended.
An issue occurring during the visualization resulting in misalignment of altLabels has been resolved.
When Data Validator is active the Taxonomy Builder feature is now disabled and a corresponding message is displayed to the user informing them that another process is currently running. When Taxonomy Builder feature is active the Data Validator cannot be launched and a corresponding message is shown to the user.
Drag & Drop functionality for a concept when creating custom relations now works as intended.
Admin users with a ReadOnly role for a specific project no longer see the edit buttons, neither can they change the project groups, nor add a note for a concept, nor apply a custom class to a concept. These users will see an error message informing them of lacking the required permissions.

Tomcat 11.0.21
GraphDB 11.3.1
Java 21
Keycloak 25.0.6
rdf4j 5.2.2
Spring Core 6.2.17
Spring Security 6.5.9
Lucene 9.9.2
Apache Spark 4.0.1
Elasticsearch 9.2.4

Warning

Please note that Graph Modeling (formerly PoolParty) uses Elasticsearch 9.2.4 which is not compatible with the previous versions. If you are using a lower version of Elasticsearch you will have to upgrade it. More details on the upgrade can be found in Elasticsearch Upgrade Guide