• Visual semantic modeling interface delivers a user-friendly environment for creating, importing, extending & editing, exploring, visualizating and documenting semantic models, based on an easy to understand, visual language. Technical and non-technical users alike can create or modify classes, relations and attributes in a visual manner
• The visual language translates to core elements of OWL and SHACL and results in a semantic model based on open modeling and validation language W3C standards
• Metadata curation and model cataloging for search and full model governance
• Tight integration between semantic models and vocabularies supports linking classes in the modeling interface to controlled vocabularies

This approach to semantic modeling does not only significantly accelerate the modeling process, but it also allows for the active involvement and participation of multiple user roles, from knowledge graph engineers to domain experts / SMEs, and business users. Additionally, it improves the quality of models and ensures early buy-in from the relevant stakeholders who are later on expected to contribute to or consume the semantic models.

More generally, using semantic modeling, organizations can make implicit information explicit by conveying context and meaning for unambiguous interpretation by both humans and machines. By following standards for interoperability, whatever has been modeled cannot only be interpreted by humans bus also queried, reasoned about, validated, serialized, exchanged and interpreted across a variety of tools and systems. This ultimately provides a foundation for reasoning that AI applications can operate with.

• Intuitive, form-based interface supports domain experts / SMEs and business users in creating and editing SKOS vocabularies to capture business-relevant terms. This helps model domain-specific knowledge in terms that business users understand and can use for analysis and in answering critical business questions.
• Support for hierarchical lists (hypernyms and hyponyms) and management of multilingual synonyms and symbols
• Performant tree visualization of and search across term hierarchies
• Ability to import vocabularies created using external tools and export vocabularies for use outside of metaphactory, simplifying communication between stakeholders
• Vocabulary cataloging for search, versioning (Git) and metadata curation
• Tight integration between ontologies and vocabularies supports linking classes in the modeling interface to controlled vocabularies

• Creation, management or import of existing dataset metadata at integration time, making such context metadata an integral part of the connected knowledge graph
• Support for DCAT and Dublin Core
• Exposure of dataset metadata in end-user oriented search interfaces, knowledge panels and custom dashboards, delivering vital context information for domain experts / SMEs and allowing them to support day-to-day tasks with traceable insights

• Ability to import or provide federated virtual access over multiple data sources
• Ability to import public ontologies and vocabularies to bootstrap semantic models and vocabularies in metaphactory. Examples of such public ontologies and vocabularies from various verticals include: the HCLSIG/PharmaOntology, the MeSH ontology, the IDMP ontology, the STW thesaurus for economics, Bibframe, schema.org, ISO15926-14, FIBO, and many more
• Option to use these assets as a basis for an own semantic model, extend them to fit specific needs, or implement them as extensions to own, proprietary knowledge models

• Easy and streamlined publishing of semantic models via API and through a Web application, supporting knowledge democratization by ensuring that everyone in the organization can access relevant models
• SSO integration to ensure compliance with company policies when it comes to user access and management
• Templating engine allows to optionally refine how information should be presented to different user groups
• Flexible, open formats ensure that semantic models can be published and shared within the community or for reuse and analysis in pre-competitive research

• Collaborative environment and streamlined modeling experience based on an agile and iterative process, allowing all relevant stakeholders – from knowledge graph engineers and taxonomists to domain experts / SMEs and business users – to equally contribute to the knowledge engineering process and collaborate on defining and continuously improving the semantic model(s), while eliminating obstacles such as external expert tools, media-breaks, or synchronization issues
• Cataloging, import/export, versioning and metadata management for semantic models, vocabularies and datasets. This improves access to assets across the organization, fosters reuse and helps to build governance processes that scale across individual projects in an organization.
• Lifecycle and change management for semantic models via a versioning mechanism and an editorial workflow that allows users to explicitly change the status of a model (from 'In development' to 'In review', etc.) lock or unlock a model for review, and communicate with other users about the changes or provide feedback about the current status
• Git integration for asset versioning, supporting the seamless embedding of knowledge modeling into governance and CI/CD processes
• Notification functionality allows to send updates to users (e.g. via email) or downstream systems on any actions or status changes related to the workflow or lifecycle of a semantic model or a vocabulary, thus enhancing the transparency and communication around asset management
• Detailed provenance documentation of an asset’s creation, owner, history and changes, made available in the visual interface, supporting traceability and helping preserve the quality of a semantic model
• Roles and permissions management enforces security, ensures that only permitted people are able to view, edit or create a semantic model, and enables accountability and traceability

• Built-in AI assists to simplify and enhance the modeling process by providing smart insights and helpful suggestions for mapping concepts to the physical data layer
• Conversational AI interface enabling users to “talk to their model” and lead contextualized conversations with the knowledge graph

• LLM-driven conversational interface allows end users to lead contextualised conversations with the graph
• The conversational interface is enabled by a range of tools and services that allow answering queries of different kinds over the knowledge graph, including the generation, refinement and execution of structured queries from natural language (NL2SPARQL) as well as answering unstructured queries using a RAG-approach (Retrieval Augmented Generation)
• Depending on the user question, the conversational interface selects and orchestrates the adequate services to answer the question and render the results in natural language or various visualization modalities using metaphactory’s built-in components (semantic tables, lists, charts, etc.)

• Semantic search caters to both explorative and targeted information needs and supports diverse search paradigms and data modalities
• Filters and facets for interactive and iterative query refinement and exploration
• Support for full-text search indices to score and rank results for responsive auto-suggestion
• Search functionality utilizing controlled and structured vocabularies for synonym search and search across narrower terms in a category
• Inclusion of sub-terms in search results when searching for broader categories
• Clipboarding functionalities to organize knowledge and support sharing and reuse of both searches and search results
• Pathfinding - Search for paths between nodes and interactively explore relations in a graph

• Rich set of components for interactive visualization and exploration of large knowledge graphs, including: interactive graphs, carousels, interactive tables, maps, charts, tree renderings and timelines
• Visualization of data along different contextual dimensions, e.g., spatial and temporal dimensions
• Browsing, filtering and sorting options available for all components
• Visualization of semantic models
• Native support for multilingual data

• Visualization of complex relations in interactive graphs
• Pathfinding - Search for paths between nodes & interactively explore relations in a graph
• In-page rich snippets of information (Knowledge Panels) allowing for contextualized exploration of individual resources
• Auto-layouts including manual refinement options and snap grids
• Ability to create interactive diagrams and add them to reports or custom applications

• Collaborative creation and editing of instance data using flexible & model-driven semantic forms
• Visual data curation interface allows end users to create, edit and connect instance data entities using the interactive graph component
• Composite inputs for semantic forms allows for the inline creation of new entities
• Rich editing components for special data types
• Auto-suggestion and validation against the knowledge graph
• Capturing of provenance information
• Automatic form setup and model-driven configuration

• Semantic Clipboard allows end users to create personalized information collections where they can store and organize named sets of knowledge graph items for reuse
• Ability to drag and drop resources from the clipboard into search interfaces or visual graphs for reuse
• Ability to save diagrams, add them to custom applications, download them as image or vector formats, and share them with other users

• Model-driven setup and configuration of metaphactory's out-of-the-box search, visualization and exploration components
• No-code wizards to identify and visually select classes, relations and attributes from the semantic model that are required to capture the end-users’ information intent
• Rendering of knowledge graph instances pertaining to the same class/type in the semantic model in a unified way, using metaphactory’s template engine
• Rich set of built-in W3C Web Components using HTML 5 syntax for composing knowledge graphs apps and dashboards declaratively
• Web-based HTML editor with syntax highlighting
• Declarative building of interactive and multilingual applications, without the need to change the source code
• Individual elements in the resulting application can be customized by templating and parameterizing them through the powerful handlebars templating language
• Customized user views depending on user roles and access permissions
• Role-based access using Single Sign-On (SSO), with, for example, OpenID Connect (OIDC), OAuth, Security Assertion Markup Language (SAML), and JSON Web Token (JWT)
• Ability to adjust dashboards to a customer’s corporate look-and-feel

Neuro-symbolic integration helps leverage the best of knowledge graphs and LLMs to drive a variety of use cases while adopting a human-in-the-loop approach. LLMs serve as an interface to the knowledge graph, enabling intelligent AI agents that understand user goals, compose and execute workflows, and learn through reinforcement or explicit instructions. This supports contextualized conversations, with responses grounded in the knowledge graph for trustworthiness, explainability and accuracy.

These AI agents are supported by a suite of tools, services, and capabilities that enhance their functionality and performance:

• Underlying semantic models with explicitly defined concepts and relations inform AI applications, ensuring trustworthiness, explainability and accuracy
• Conversational interfaces enable users to lead contextualized conversations with the graph using natural language for insights, knowledge discovery or semantic modeling
• AI tools and services facilitate a variety of query-answering paradigms:
• Generation, refinement and execution of structured queries from natural language (NL2SPARQL)
• Answering unstructured queries using a RAG approach (Retrieval Augmented Generation) and supporting LLMs with access to factual knowledge
• Bridging to additional services like pathfinding, similarity searches and other ML algorithms
• Knowledge consumption is driven by data context and user intent, incorporating verbalization, summarization and visualizations like tables, charts and timelines

With these capabilities in place, the integration supports a range of use cases, including:

• Search and analytics using intuitive interaction patterns such as a conversational interface
• Knowledge generation by extracting structured knowledge (entities, facts, relations) from unstructured sources and contextualizing and enriching text with structured knowledge
• Semantic model extension by identifying new types of concepts and relations using LLMs

• Repository Management - Vendor-independent proxy over multiple SPARQL 1.1 compliant data repositories including access control
• Multiple W3C standardized HTTP data access interfaces with fine grained access control levels: SPARQL 1.1 Protocol, SPARQL 1.1 Graph Store HTTP Protocol, Linked Data Platform 1.0, RDF Publishing / Content Negotiation for individual resources

• Hybrid federation services over heterogeneous data sources using the federation engine Ephedra
• Transparent federation of multiple SPARQL endpoints under a single virtual endpoint using the federation engine FedX

• Query as a Service
• Ability to template queries and expose them during runtime as dynamic and parameterisable REST APIs for 3rd party application integration (e.g., Workflow Pipeline, Knime, R, etc.) with configurable output formats (JSON, CSV, RDF)
• Controlled access to published query results using access permissions, data hiding, and encapsulation
• Label service - ubiquitous access to human readable labels for knowledge graph resources
• Description service - provides contextual information for a knowledge graph entity when exploring and searching data, e.g., for disambiguation purpose
• Lookup service - provides an abstraction layer over keyword indices
• Role-based access – all service endpoints are protected by the role-based access model and single sign-on (OAuth2, OpenId Connect, SAML2, LDAP)

• One comprehensive solution for connecting, discovering and managing data across physical and virtual data repositories (or named graphs)
• Support for data ingestion
• Standard connectors for a variety of data formats
• Low level data profiling, i.e., inspection of data types, data partitioning, statistics
• Centralized management of namespaces and prefixes
• SPARQL Query Editor: Contextualization, syntax highlighting, and auto-completion
• Query Catalog: Ability to save and restore queries

• SHACL-based Rule Engine: Support for native SHACL rules to establish technical conformity of data
• Integration with database native SHACL validation engines
• Ability to define rules based on SPARQL queries to run checks and validations against custom business logic
• Charts displaying high-level, visual aggregations are used to quickly assess current data quality status and trends over time, including previous data quality reports
• Detailed violation reports allow for inspecting and drilling down to individual violations, e.g., to identify the knowledge graph resources violating a rule