> Details

Baldazzi, Teodoro [Author]; Bellomarini, Luigi [Author]; Ceri, Stefano [Author]; Colombo, Andrea [Author]; Gentili, Andrea [Author]; Sallinger, Emanuel [Author]; Atzeni, Paolo [Author] ; Teodoro Baldazzi and Luigi Bellomarini and Stefano Ceri and Andrea Colombo and Andrea Gentili and Emanuel Sallinger and Paolo Atzeni [Contributor]

Explaining Enterprise Knowledge Graphs with Large Language Models and Ontological Reasoning

Reference
management

Bookmarks

Remove from
bookmarks

Share this on Whatsapp

Close

Bookmarks



You can manage bookmarks using lists, please log in to your user account for this.
  • Media type: Text; E-Article; Electronic Conference Proceeding
  • Title: Explaining Enterprise Knowledge Graphs with Large Language Models and Ontological Reasoning
  • Contributor: Baldazzi, Teodoro [Author]; Bellomarini, Luigi [Author]; Ceri, Stefano [Author]; Colombo, Andrea [Author]; Gentili, Andrea [Author]; Sallinger, Emanuel [Author]; Atzeni, Paolo [Author]
  • Published: Schloss Dagstuhl – Leibniz-Zentrum für Informatik, 2024
  • Language: English
  • DOI: https://doi.org/10.4230/OASIcs.Tannen.1
  • Keywords: knowledge graphs ; provenance ; ontological reasoning ; language models
  • Origination:
  • Footnote: Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
  • Description: In recent times, the demand for transparency and accountability in AI-driven decisions has intensified, particularly in high-stakes domains like finance and bio-medicine. This focus on the provenance of AI-generated conclusions underscores the need for decision-making processes that are not only transparent but also readily interpretable by humans, to built trust of both users and stakeholders. In this context, the integration of state-of-the-art Large Language Models (LLMs) with logic-oriented Enterprise Knowledge Graphs (EKGs) and the broader scope of Knowledge Representation and Reasoning (KRR) methodologies is currently at the cutting edge of industrial and academic research across numerous data-intensive areas. Indeed, such a synergy is paramount as LLMs bring a layer of adaptability and human-centric understanding that complements the structured insights of EKGs. Conversely, the central role of ontological reasoning is to capture the domain knowledge, accurately handling complex tasks over a given realm of interest, and to infuse the process with transparency and a clear provenance-based explanation of the conclusions drawn, addressing the fundamental challenge of LLMs' inherent opacity and fostering trust and accountability in AI applications. In this paper, we propose a novel neuro-symbolic framework that leverages the underpinnings of provenance in ontological reasoning to enhance state-of-the-art LLMs with domain awareness and explainability, enabling them to act as natural language interfaces to EKGs.
  • Access State: Open Access