The Moral Foundations Reddit Corpus
This provides a domain-specific resource for NLP and social science researchers to improve understanding of moral rhetoric, though it is incremental as it extends existing annotation efforts to a new platform.
The authors tackled the lack of large, hand-annotated datasets for detecting moral sentiment in text by creating the Moral Foundations Reddit Corpus, a collection of 16,123 Reddit comments annotated for 8 moral categories, and found that large language models like Llama3-8B and Ministral-8B lag behind fine-tuned encoders like BERT in performance on this task.
Moral framing and sentiment can affect a variety of online and offline behaviors, including donation, environmental action, political engagement, and protest. Various computational methods in Natural Language Processing (NLP) have been used to detect moral sentiment from textual data, but achieving strong performance in such subjective tasks requires large, hand-annotated datasets. Previous corpora annotated for moral sentiment have proven valuable, and have generated new insights both within NLP and across the social sciences, but have been limited to Twitter. To facilitate improving our understanding of the role of moral rhetoric, we present the Moral Foundations Reddit Corpus, a collection of 16,123 English Reddit comments that have been curated from 12 distinct subreddits, hand-annotated by at least three trained annotators for 8 categories of moral sentiment (i.e., Care, Proportionality, Equality, Purity, Authority, Loyalty, Thin Morality, Implicit/Explicit Morality) based on the updated Moral Foundations Theory (MFT) framework. We evaluate baselines using large language models (Llama3-8B, Ministral-8B) in zero-shot, few-shot, and PEFT settings, comparing their performance to fine-tuned encoder-only models like BERT. The results show that LLMs continue to lag behind fine-tuned encoders on this subjective task, underscoring the ongoing need for human-annotated moral corpora for AI alignment evaluation. Keywords: moral sentiment annotation, moral values, moral foundations theory, multi-label text classification, large language models, benchmark dataset, evaluation and alignment resource