OpenDialKG: Explainable Conversational Reasoning with Attention-based Walks over Knowledge Graphs

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 845–854 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 845 OpenDialKG: Explainable Conversational Reasoning with Attention-based Walks over Knowledge Graphs Seungwhan Moon, Pararth Shah, Anuj Kumar, Rajen Subba Facebook Conversational AI {shanemoon, pararths, anujk, rasubba@}fb.com Abstract We study a conversational reasoning model that strategically traverses through a large- scale common fact knowledge graph (KG) to introduce engaging and contextually diverse entities and attributes. For this study, we col- lect a new Open-ended Dialog ↔ KG par- allel corpus called OpenDialKG, where each utterance from 15K human-to-human role- playing dialogs is manually annotated with ground-truth reference to corresponding enti- ties and paths from a large-scale KG with 1M+ facts. We then propose the DialKG Walker model that learns the symbolic transitions of dialog contexts as structured traversals over KG, and predicts natural entities to introduce given previous dialog contexts via a novel domain-agnostic, attention-based graph path decoder. Automatic and human evaluations show that our model can retrieve more natu- ral and human-like responses than the state-of- the-art baselines or rule-based models, in both in-domain and cross-domain tasks. The pro- posed model also generates a KG walk path for each entity retrieved, providing a natural way to explain conversational reasoning. 1 Introduction The key element of an open-ended dialog sys- tem is its ability to understand conversational con- texts and to respond naturally by introducing rele- vant entities and attributes, which often leads to increased engagement and coherent interactions (Chen et al., 2018). While a large-scale knowledge graph (KG) includes vast knowledge of all the re- lated entities connected via one or more factual connections from conversational contexts, the core challenge is in the domain-agnostic and scalable prediction of a small subset from those reachable entities that follows natural conceptual threads that can keep conversations engaging and meaningful. Hence, we study a data-driven reasoning model Figure 1: Conversational reasoning with a parallel (a) dialog and (b) knowledge graph (KG) corpus. Diverse topical jumps across open-ended multi-turn dialogs are annotated and grounded with a large-scale common- fact KG. To generate a KG entity response at each di- alog turn, the model learns walkable paths within KG that lead to engaging and natural topics or entities given dialog context, while pruning non-ideal (albeit factu- ally correct) KG paths among 1M+ candidate facts. that map dialog transitions with KG paths, aimed at identifying a subset of ideal entities to mention as a response to previous dialog contexts. Figure 1 illustrates a motivating dialog exam- ple between two conversation participants, which spans multiple related KG entities from a start- ing seed entity The Catcher in the Rye. Specif- ically, we observe that there exists a small subset of walkable patterns within a KG or a preferred se- quence of graph traversal steps which often leads to more engaging entities or attributes than oth- ers (e.g. Literacy Realism, Nathaniel Hawthorne, etc. vs. Catch Me If You Can, 277, etc. …

Content truncated to 500 words