Automated detection of patronizing language in english and mandarin
Patronizing and Condescending Language Detection in English and Mandarin
Artificial Intelligence
Summary
Patronizing and condescending language is a subtle form of harmful speech that can be hard to spot, even for people. The authors developed computer programs that try to recognize this kind of language automatically in English and Mandarin text. They tested two methods: one using fixed word meanings and a simpler model, and another using an advanced language understanding model called BERT. The BERT-based method worked better, although it still has room for improvement.
What this means in practice
- •For social media platform teams: Detect patronizing and condescending comments automatically to improve content moderation and support vulnerable groups.
- •For customer service centers: Flag patronizing language in customer feedback to train staff for better communication and reduce bias.
Authors
Xinyi Vicky Xiang, Catherine Qingyun Wang, Paula Cortés Lemos
Abstract
Patronizing and condescending language (PCL) is a form of harmful language that belittles and assumes pity for vulnerable groups, reinforcing negative societal stereotypes. However, unlike overt hate speech, it is often difficult to detect, even for human judges. This paper proposes a system for the automated detection of PCL. Our first approach uses static word embeddings and a convolutional neural network for classification, while our second approach uses pre-trained BERT embeddings fine-tuned for the PCL task. Our second approach was considerably more successful, and achieved an F1 score of 0.528 over the development set.