Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WEBDial, a Multi-domain, Multitask Statistical Dialogue Framework with RDF

Morgan Veyret, Jean-Baptiste Duchene, Kekeli Afonouvi, Quentin Brabant, Gwenole Lecorve, Lina M. Rojas-Barahona

TL;DR

Slot-value representations constrain expressivity, scalability, and explainability in dialogue systems. This paper proposes WEBDial, a modular framework that uses RDF graphs anchored to a separated ontology to represent dialogue knowledge, tasks, and domains. The authors implement and evaluate four applications across increasing complexity, illustrating how RDF-based semantics can support cross-domain information seeking, booking, and complex tasks. The approach promises improved expressivity, easier domain expansion, and potential for more explainable, causally grounded dialogue decisions in real-world settings.

Abstract

Typically available dialogue frameworks have adopted a semantic representation based on dialogue-acts and slot-value pairs. Despite its simplicity, this representation has disadvantages such as the lack of expressivity, scalability and explainability. We present WEBDial: a dialogue framework that relies on a graph formalism by using RDF triples instead of slot-value pairs. We describe its overall architecture and the graph-based semantic representation. We show its applicability from simple to complex applications, by varying the complexity of domains and tasks: from single domain and tasks to multiple domains and complex tasks.

WEBDial, a Multi-domain, Multitask Statistical Dialogue Framework with RDF

TL;DR

Slot-value representations constrain expressivity, scalability, and explainability in dialogue systems. This paper proposes WEBDial, a modular framework that uses RDF graphs anchored to a separated ontology to represent dialogue knowledge, tasks, and domains. The authors implement and evaluate four applications across increasing complexity, illustrating how RDF-based semantics can support cross-domain information seeking, booking, and complex tasks. The approach promises improved expressivity, easier domain expansion, and potential for more explainable, causally grounded dialogue decisions in real-world settings.

Abstract

Typically available dialogue frameworks have adopted a semantic representation based on dialogue-acts and slot-value pairs. Despite its simplicity, this representation has disadvantages such as the lack of expressivity, scalability and explainability. We present WEBDial: a dialogue framework that relies on a graph formalism by using RDF triples instead of slot-value pairs. We describe its overall architecture and the graph-based semantic representation. We show its applicability from simple to complex applications, by varying the complexity of domains and tasks: from single domain and tasks to multiple domains and complex tasks.
Paper Structure (11 sections, 5 figures)

This paper contains 11 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Architectural Description
  4. NLU
  5. DST
  6. DP
  7. NLG
  8. Ontology
  9. Semantic Representation
  10. Applications
  11. Conclusion and Future Work

Figures (5)

  • Figure 1: High-level depiction of the proposed dialogue framework. Arrows indicate data flow.
  • Figure 2: Excerpt of the ontology of Multiwoz.
  • Figure 3: Semantics for the user's input: "I no longer have video on-demand. The screen displays unavailable service. It also displays error codes PO1 and D24". The dialogue act is inform.
  • Figure 4: Semantics for the system's instruction: "Press the button on the front of the decoder for approximately 5 seconds then the configuration screen of your decoder is displayed". The dialogue act is command.
  • Figure 5: Excerpt of the grammar for the proposed meaning representation