Neural Generation Meets Real People: Towards Emotionally Engaging Mixed-Initiative Conversations

By Ashwin Paranjape, Abigail See, Kathleen Kenealy, Haojun Li, Amelia Hardy, Peng Qi, Kaushik Ram Sadagopan, Nguyet Minh Phu, Dilara Soylu, Christopher D. Manning. 2020

42 citations recorded by the source corpus.

We present Chirpy Cardinal, an open-domain dialogue agent, as a research platform for the 2019 Alexa Prize competition. Building an open-domain socialbot that talks to real people is challenging - such a system must meet multiple user expectations such as broad world knowledge, conversational style, and emotional connection. Our socialbot engages users on their terms - prioritizing their interests, feelings and autonomy. As a result, our socialbot provides a responsive, personalized user experience, capable of talking knowledgeably about a wide variety of topics, as well as chatting empathetically about ordinary life. Neural generation plays a key role in achieving these goals, providing the backbone for our conversational and emotional tone. At the end of the competition, Chirpy Cardinal progressed to the finals with an average rating of 3.6/5.0, a median conversation duration of 2 minutes 16 seconds, and a 90th percentile duration of over 12 minutes.

Subfields: Artificial Intelligence, Developmental and Educational Psychology, Social Psychology, Natural Language Processing, LLM Agents & Tool Use

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Developmental and Educational Psychology

In Social Psychology

In Natural Language Processing

In LLM Agents & Tool Use

Jason Aaron Edward Weston

Citations in the corpus

Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling (2021, background): “Paranjape et al. (2020) measure users’ re-offense rates following different response strategies, ﬁnding avoidance to be the most successful approach by this metric.”
Recipes for Safety in Open-domain Chatbots (2020, methodology): “In this work, we consider two approaches, which we refer to as safe responses, and non sequiturs (Curry and Rieser, 2019; Paranjape et al., 2020).”
Hindsight: Posterior-guided training of retrievers for improved open-ended generation (2021, background): “Further, when a “trusting” generator is deployed in real-life settings (e.g., in open-domain socialbots; Paranjape et al. 2020), external business logic can select an appropriate passage from the top-k retrieved passages and effectively control the generated content.”
Bot-Adversarial Dialogue for Safe Conversational Agents (2021, background): “While several different “safe” response strategies can be considered (Curry and Rieser, 2019; Paranjape et al., 2020), in this work we respond with a non-sequitur: we select a topic at random from 1,087 topics judged as safe from the Wizard of Wikipedia conversational topic list (Dinan et al.,…”
SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures (2021, background): “Paranjape et al. (2020) look at re-offense rates after various response types.”
ConvAbuse: Data, Analysis, and Benchmarks for Nuanced Abuse Detection in Conversational AI (2021, background): “These include several submissions to the Amazon Alexa Challenge 2 (Cercas Curry et al., 2018; Kha-tri et al., 2018; Paranjape et al., 2020).”
Automatically Exposing Problems with Neural Dialog Models (2021, background): “Compared to modularized dialog systems which are designed to avoid these problems (Yu et al., 2019; Paranjape et al., 2020), ﬁxing these issues with end-to-end neural models is more challenging, which may hinder real world use of trained models (Wolf et al., 2017; Simonite, 2021).”
Further Advances in Open Domain Dialog Systems in the Third Alexa Prize Socialbot Grand Challenge (2020, methodology): “For DA and intent classification, Gunrock used active learning to improve model performance; Athena expanded the DA ontology used by Gunrock, with more types for user questions, and used an ensemble model for DA classification; Alquist also introduced hierarchical DA classes; Chirpy Cardinal (Paranj”