Computational Politeness in Natural Language Processing: A Survey

TL;DR

This survey reviews computational methods for detecting and generating politeness in text, highlighting key milestones, datasets, approaches, and future directions in natural language processing.

Contribution

It compiles and analyzes past research on computational politeness, emphasizing four major milestones and providing a comprehensive overview of datasets, techniques, and trends.

Findings

Identification of key features for politeness detection

Inclusion of contextual and social factors in models

Summary of datasets and performance benchmarks

Abstract

Computational approach to politeness is the task of automatically predicting and generating politeness in text. This is a pivotal task for conversational analysis, given the ubiquity and challenges of politeness in interactions. The computational approach to politeness has witnessed great interest from the conversational analysis community. This article is a compilation of past works in computational politeness in natural language processing. We view four milestones in the research so far, viz. supervised and weakly-supervised feature extraction to identify and induce politeness in a given text, incorporation of context beyond the target text, study of politeness across different social factors, and study the relationship between politeness and various sociolinguistic cues. In this article, we describe the datasets, approaches, trends, and issues in computational politeness research. We also discuss representative performance values and provide pointers to future works, as given in the prior works. In terms of resources to understand the state-of-the-art, this survey presents several valuable illustrations, most prominently, a table summarizing the past papers along different dimensions, such as the types of features, annotation techniques, and datasets used.