Machine Learning-based Guilt Detection in Text
We introduce a new Natural Language Processing (NLP), task called guilt detection. This task focuses on detecting guilty in text. We recognize guilt as an important emotion, which has never been studied before in NLP. Our goal is to provide a finer-grained analysis. We created VIC to address the lack publicly available corpora that are suitable for guilt detection. This dataset contains 4,622 texts from existing emotion detection datasets, which we binarized in guilt and no-guilt categories. The highest performing model achieved a 72% F1 score using traditional machine learning techniques. We used bag-of words and term frequency-inverse documents frequency features. Our study is a first step in understanding guilt in texts and opens up the possibility of future research.
Source:
https://www.nature.com/articles/s41598-023-38171-0