This repo houses the code submitted to SemEval 2023 for Task 4: Human Value Detection. The peer-reviewed paper associated with this repo can be found here.
There are four models within this repository, a supervised XGBoost, two supervised Ensemble, and an unsupervised Threshold Comparison model.
Executing the main.py script will run all four models on the training and validation data. Both bert-base-uncased and all-MiniLM-L6-v2 were
used to generate embeddings for the data provided by the task organizers. The hyperparameters have been tuned to the F1-score evaluation metric
used in the task.