PRAGEU: Praise Generation Using End-to-End Neural Model

Anonymous

Abstract: Humans employ a range of conversation strategies during conversation to achieve multiple goals in dialogue. One of such strategies is a positive evaluation made by a person of another's attributes, known as \textit{Praise}. State of the art neural dialogue models attempt to engage human users without taking into account this conversation strategy. Hence in this work, we present a method of generating praise using state of the art natural language generation models. We achieve this by collecting a dataset using amazon mechanical turk (AMT) using Persona-Chat and create a new corpus called Praise-on-Persona (POP) and fine-tune various models to generate praise. Our results show that large language models can learn to link an attribute and a praise associated with it, such a a \textit{Professor} and \textit{Intelligence} or \textit{PhD} and \textit{hard work}.

Paper Type: short

Research Area: Generation

Contribution Types: NLP engineering experiment

Languages Studied: English

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.