Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Using Domain Knowledge to Guide Dialog Structure Induction via Neural Probabilistic Soft Logic

Connor Pryor, Quan Yuan, Jeremiah Liu, Mehran Kazemi, Deepak Ramachandran, Tania Bedrax-Weiss, Lise Getoor

TL;DR

Over three dialog structure induction datasets and across unsupervised and semi-supervised settings for standard and cross-domain generalization, the injection of symbolic knowledge using NEUPSL DSI provides a consistent boost in performance over the canonical baselines.

Abstract

Dialog Structure Induction (DSI) is the task of inferring the latent dialog structure (i.e., a set of dialog states and their temporal transitions) of a given goal-oriented dialog. It is a critical component for modern dialog system design and discourse analysis. Existing DSI approaches are often purely data-driven, deploy models that infer latent states without access to domain knowledge, underperform when the training corpus is limited/noisy, or have difficulty when test dialogs exhibit distributional shifts from the training domain. This work explores a neural-symbolic approach as a potential solution to these problems. We introduce Neural Probabilistic Soft Logic Dialogue Structure Induction (NEUPSL DSI), a principled approach that injects symbolic knowledge into the latent space of a generative neural model. We conduct a thorough empirical investigation on the effect of NEUPSL DSI learning on hidden representation quality, few-shot learning, and out-of-domain generalization performance. Over three dialog structure induction datasets and across unsupervised and semi-supervised settings for standard and cross-domain generalization, the injection of symbolic knowledge using NEUPSL DSI provides a consistent boost in performance over the canonical baselines.

Using Domain Knowledge to Guide Dialog Structure Induction via Neural Probabilistic Soft Logic

TL;DR

Over three dialog structure induction datasets and across unsupervised and semi-supervised settings for standard and cross-domain generalization, the injection of symbolic knowledge using NEUPSL DSI provides a consistent boost in performance over the canonical baselines.

Abstract

Dialog Structure Induction (DSI) is the task of inferring the latent dialog structure (i.e., a set of dialog states and their temporal transitions) of a given goal-oriented dialog. It is a critical component for modern dialog system design and discourse analysis. Existing DSI approaches are often purely data-driven, deploy models that infer latent states without access to domain knowledge, underperform when the training corpus is limited/noisy, or have difficulty when test dialogs exhibit distributional shifts from the training domain. This work explores a neural-symbolic approach as a potential solution to these problems. We introduce Neural Probabilistic Soft Logic Dialogue Structure Induction (NEUPSL DSI), a principled approach that injects symbolic knowledge into the latent space of a generative neural model. We conduct a thorough empirical investigation on the effect of NEUPSL DSI learning on hidden representation quality, few-shot learning, and out-of-domain generalization performance. Over three dialog structure induction datasets and across unsupervised and semi-supervised settings for standard and cross-domain generalization, the injection of symbolic knowledge using NEUPSL DSI provides a consistent boost in performance over the canonical baselines.
Paper Structure (36 sections, 26 equations, 12 figures, 7 tables)

This paper contains 36 sections, 26 equations, 12 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background
  4. Direct Discrete Variational Recurrent Neural Networks
  5. Probabilistic Soft Logic
  6. Neural Probabilistic Soft Logic Dialogue Structure Induction
  7. Problem Formulation
  8. Integrating Neural and Symbolic Learning under NeuPSL DSI
  9. Improving soft logic constraints for gradient learning
  10. Stronger control of posterior collapse via weighted bag of words
  11. Experimental Evaluation
  12. Datasets
  13. Constraints
  14. Metrics and Methodology
  15. Main Results
  16. ...and 21 more sections

Figures (12)

  • Figure 1: Example dialog structure for the goal-oriented task booking a hotel.
  • Figure 2: The high-level pipeline of the NeuPSL DSI$\;$ learning procedure.
  • Figure 3: Average AMI for MultiWoZ, SGD Synthetic, and SGD Real (Standard Generalization, Domain Generalization, and Domain Adaptation) on three constrained few-shot settings: 1-shot, proportional 1-shot, and 3-shot. Hidden representation learning graphs are included in the Appendix.
  • Figure 4: Average AMI performance for SGD Real (Standard Generalization, Domain Generalization, and Domain Adaptation) on three highly constrained few-shot settings: 1 shot, proportional 1 shot, and 3 shot.
  • Figure 5: SGD Structure Induction Constraint Model
  • ...and 7 more figures