Human Value Identification Project

This project focuses on the implementation of a BERT-based model to address the multi-label classification task of detecting human values in textual arguments. Utilizing the dataset from the "Human Value Detection 2023" challenge, our approach involves fine-tuning a RoBERTa-based model to classify arguments into categories derived from social science literature. Three models were developed, considering different parts of the arguments: conclusion, premise, and stance. The best model achieved a macro F1-score of 0.746, demonstrating that incorporating the premise significantly enhances performance. This repository contains all the necessary code and documentation to reproduce our experiments and results. Additional information regarding the results obtained and the specifications on the techniques used can be found in the file human_value_identification_report.pdf