Sharif-STR at SemEval-2024 Task 1: Transformer as a Regression Model for Fine-Grained Scoring of Textual Semantic Relations
This repository contains code and resources for computing the relatedness score between two samples using a transformer-based regression model. The primary purpose of this notebook is to apply a correlation metric on a test set, evaluating the model's success in measuring the relatedness between pairs of samples.
The notebook is structured into several parts, with a focus on using a transformer-based regression model, specifically fine-tuning the BERT model (though other models can be used as well). There is also a section to augment the dataset with a T5-paraphraser.
The notebook.ipynb script is dedicated to fine-tuning the BERT model for the relatedness scoring task. This part includes steps for:
- Loading and preprocessing the dataset.
- Configuring and fine-tuning the transformer model.
- Evaluating the model on a test set.
- Extracting and saving relatedness scores.
- Generating dataset with the T5-paraphraser
If you find this work valuable for your research, please consider citing the relevant paper or source for the transformer model used.
This project is licensed under the MIT License - see the LICENSE file for details.
We would like to acknowledge the developers of the RoBERTa model and SemEval organizers for providing the datasets that contribute to this research.