Interview: Large-scale Modeling of Media Dialog with Discourse Patterns and Knowledge Grounding

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 8129–8141, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 8129 Interview: Large-scale Modeling of Media Dialog with Discourse Patterns and Knowledge Grounding Bodhisattwa Prasad Majumder∗ Shuyang Li∗ Jianmo Ni† Julian McAuley Computer Science and Engineering University of California, San Diego {bmajumde, shl008, jin018, jmcauley}@ucsd.edu Abstract In this work, we perform the first large-scale analysis of discourse in media dialog and its impact on generative modeling of dialog turns, with a focus on interrogative patterns and use of external knowledge. Discourse analy- sis can help us understand modes of persua- sion, entertainment, and information elicita- tion in such settings, but has been limited to manual review of small corpora. We intro- duce Interview—a large-scale (105K conver- sations) media dialog dataset collected from news interview transcripts—which allows us to investigate such patterns at scale. We present a dialog model that leverages exter- nal knowledge as well as dialog acts via aux- iliary losses and demonstrate that our model quantitatively and qualitatively outperforms strong discourse-agnostic baselines for dialog modeling—generating more specific and topi- cal responses in interview-style conversations. 1 Introduction Much of the news, information, and punditry the general public listens to and reads consists of media dialog—a category of open-domain conversations between an interviewer and interviewee centered on world events and situational context. A system for modeling media dialog from the perspective of one of these roles can help us better understand how media persuades and informs the public (Southwell et al., 2018). Thus, while recent work in dialog modeling has focused on goal-oriented (Bordes et al., 2017), spontaneous (Shao et al., 2017), or synthetic open-domain chit-chat (Li et al., 2017; Dinan et al., 2019; Gopalakrishnan et al., 2019), we aim to analyze discourse patterns in media dialog and their impact on dialog modeling. Media dialog differs linguistically and in pur- pose from unstructured, spontaneous conversation ∗ denotes equal contribution † Now at Google Figure 1: Our dialog model incorporates grounding documents alongside dialog history. We also leverage the dialog patterns and interrogative positioning by the host via auxiliary losses. such as open-domain chitchat, and both the topical content and interlocutor intent are heavily influ- enced by the social, cultural, and temporal setting (Weizman, 2008). The study of media dialog has traditionally focused on individual and manual re- view of small-scale (<200K word) news corpora (Bednarek, 2006; van Dijk, 2011), and we see an opportunity to scale some forms of discourse anal- ysis to tens of thousands of such documents. In this work, we perform the first large-scale automatic analysis of structural components (response-type patterns) and question type categorization on me- dia dialog, specifically for English news interviews. We show that predicting discourse features can improve generative dialog modeling performance, demonstrating the degree to which discourse struc- ture impacts an interviewer’s choice of response type and content. News interviews are also heav- ily situation-grounded and contextualized by past events and world knowledge. We explore methods to associate each conversation …

Content truncated to 500 words