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

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 7631–7652 July 9-14, 2023 ©2023 Association for Computational Linguistics Using Domain Knowledge to Guide Dialog Structure Induction via Neural Probabilistic Soft Logic Connor Pryor ∗ UC Santa Cruz cfpryor@ucsc.edu Quan Yuan Google Research yquan@google.com Jeremiah Liu Google Research Harvard University jereliu@google.com Mehran Kazemi Google Research mehrankazemi@ google.com Deepak Ramachandran Google Research ramachandrand@google.com Tania Bedrax-Weiss Google Research tbedrax@google.com Lise Getoor UC Santa Cruz getoor@ucsc.edu Abstract Dialog Structure Induction (DSI) is the task of inferring the latent dialog structure (i.e., a set of dialog states and their temporal transi- tions) of a given goal-oriented dialog. It is a critical component for modern dialog system design and discourse analysis. Existing DSI ap- proaches 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 diffi- culty when test dialogs exhibit distributional shifts from the training domain. This work ex- plores a neural-symbolic approach as a poten- tial solution to these problems. We introduce Neural Probabilistic Soft Logic Dialogue Struc- ture 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 hid- den 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 general- ization, the injection of symbolic knowledge us- ing NEUPSL DSI provides a consistent boost in performance over the canonical baselines. 1 Introduction The seamless integration of prior domain knowl- edge into the neural learning of language structure has been an open challenge in the machine learn- ing and natural language processing communities. In this work, we inject symbolic knowledge into the neural learning process of a two-party dialog structure induction (DSI) task (Zhai and Williams, 2014; Shi et al., 2019). This task aims to learn a graph, known as the dialog structure, capturing ∗Work done during internship at Google. Figure 1: Example dialog structure for the goal-oriented task booking a hotel. the potential flow of states occurring in a dialog dataset for a specific task-oriented domain, e.g., Figure 1 represents a possible dialog structure for the goal-oriented task of booking a hotel. Nodes in the dialog structure represent conversational topics or dialog acts that abstract the intent of individual utterances, and edges represent transitions between dialog acts over successive turns of the dialog. Similar to the motivation described in (Shi et al., 2019), previous work in DSI has been split between supervised and unsupervised methods. In particu- lar, traditional supervised methods (Jurafsky, 1997) relied on dialog structure hand-crafted by human domain experts. Unfortunately, this process is labor-intensive and, in most situations, does not generalize easily to new domains. Therefore, recent work attempts to overcome this limitation by study- ing unsupervised DSI; e.g., hidden Markov models (Chotimongkol, 2008; lan Ritter et al., 2010; Zhai …

Content truncated to 500 words