Research

Atzeni, Baldazzi, Bellomarini, Laurenza, Sallinger — A framework combining semantic awareness with LLMs for accurate natural language query answering over enterprise knowledge graphs, grounding responses in ontological reasoning.

Colombo, Baldazzi, Bellomarini, Sallinger, Ceri — Introduces a template-driven approach for generating human-readable explanations of AI inference over financial knowledge graphs, enabling transparency in high-stakes decision-making.

Benedetto, Calautti, Hammad, Sallinger, Vlad — Presents an efficient algorithm for rewriting queries over warded ontologies into Datalog, enabling scalable and complete reasoning without exponential blowup.

Baldazzi, Bellomarini, Sallinger — Demonstrates the Vadalog reasoning system applied to complex financial scenarios, leveraging ontological rules to automate regulatory compliance and risk analysis at scale.

Baldazzi, Bellomarini, Sallinger — Explores how knowledge graphs enhance LLM reasoning through neurosymbolic integration, bridging declarative domain knowledge with neural language understanding.

Bellomarini, Benedetto, Brandetti, Sallinger, Vlad — Presents Vadalog Parallel, a distributed system for large-scale ontological reasoning with Datalog+/-, delivering high-performance execution across enterprise-scale knowledge graphs.

Baldazzi, Bellomarini, Ceri, Colombo, Gentili, Sallinger, Atzeni — A neuro-symbolic framework using ontological reasoning provenance to make LLMs domain-aware and explainable, enabling natural language interfaces to enterprise knowledge graphs.

Baldazzi, Bellomarini, Ceri, Colombo, Gentili, Sallinger — An interactive explanation framework for Datalog-based reasoning in Vadalog, letting business users query why specific conclusions were derived from enterprise knowledge graphs.

Baldazzi, Bellomarini, Favorito, Sallinger — Extends ontological reasoning to streaming architectures using Shy and Warded Datalog+/-, enabling continuous real-time knowledge graph updates over high-velocity data streams.

Baldazzi, Benedetto, Bellomarini, Sallinger, Vlad — Proposes a hybrid approach combining the rigour of ontological reasoning with the flexibility of LLMs to handle incomplete or inconsistent enterprise knowledge in practice.

Colombo, Baldazzi, Bellomarini, Gentili, Sallinger — Applies LLMs to automatically generate DatalogMTL temporal logic programs for modelling MiCAR-compliant cryptocurrency market regulations.

Baldazzi, Bellomarini, Ceri, Colombo, Gentili, Sallinger — A neurosymbolic architecture using ontological reasoning to construct task-specific training corpora for fine-tuning LLMs, combining structured domain knowledge with neural adaptability.

Dwyer, Baldazzi, Davies, Sallinger, Vlad — Demonstrates Vadalog's application to healthcare, using rule-based ontological reasoning over electronic health records to model and analyse complex patient care pathways.

Baldazzi, Bellomarini, Sallinger — Showcases the Vadalog system's capabilities for reasoning over complex financial scenarios, from regulatory compliance to risk assessment using knowledge graph technology.

Baldazzi, Bellomarini, Gerschberger, Jami, Magnanimi, Nissl, Pavlovic, Sallinger — A comprehensive overview of the Vadalog system's architecture, extensions, and real-world deployments across finance, healthcare, and enterprise knowledge management.

Bellomarini, Benedetto, Brandetti, Sallinger — Introduces "Harmless EGDs," a tractable class of equality-generating dependencies enabling multi-criteria clustering, data fusion, and graph traversal within Warded Datalog+/-.

Bellomarini, Bencivelli, Biancotti, Blasi, Benedetto et al. — Applies knowledge graph reasoning to analyse corporate takeover tactics and strategies, combining ontological rules with financial data to detect beneficial ownership patterns.

Bellomarini, Benedetto, Gottlob, Sallinger — The definitive reference for the Vadalog architecture—a system combining Datalog+/- with advanced reasoning capabilities for automated knowledge graph processing at enterprise scale.

Baldazzi, Bellomarini, Favorito, Sallinger — Establishes formal connections between Shy and Warded Datalog+/- fragments, clarifying their expressive power and computational properties for ontological reasoning.

Atzeni, Baldazzi, Bellomarini, Sallinger — Presents iWarded, a versatile benchmark generator for Warded Datalog+/- reasoning systems, enabling systematic and reproducible performance evaluation of ontological reasoning engines.

Bellomarini, Benedetto, Sallinger — An accessible overview of Vadalog's architecture, capabilities, and real-world applications for automated reasoning over large-scale enterprise knowledge graphs.

Baldazzi, Benedetto, Brandetti, Vlad, Bellomarini, Sallinger — Extends Datalog-based reasoning with heuristic mechanisms to handle uncertainty and incompleteness in enterprise knowledge graphs, improving practical performance and applicability.

Baldazzi, Benedetto, Brandetti, Vlad, Bellomarini — Applies heuristic reasoning techniques to financial knowledge graphs, enabling fast approximate inference for regulatory compliance and financial risk detection.

Vlad, Vahdati, Nayyeri, Bellomarini, Sallinger — Explores hybrid reasoning combining symbolic logic with neural embeddings for financial knowledge graphs, balancing logical explainability with representational flexibility.

Baldazzi, Atzeni — Investigates the challenges of applying Warded Datalog+/- to financial knowledge graphs in the presence of harmful joins, proposing solutions for tractable reasoning in complex regulatory settings.

Atzeni, Bellomarini, Iezzi, Sallinger, Vlad — Presents techniques for constructing and reasoning over company ownership knowledge graphs, automating the discovery of ultimate beneficial owners from heterogeneous distributed data sources.

Atzeni, Bellomarini, Iezzi, Sallinger, Vlad — Explores augmenting logic-based knowledge graphs with additional embedding-derived information for company ownership analysis, improving completeness of enterprise data.

Atzeni, Bellomarini, Benedetto, Sallinger — Examines join-based traversal strategies for knowledge graphs within the Vadalog framework, studying the balance between expressive power and computational efficiency.