dataset-pruning-sarcasm-detection

URL: https://github.com/priyank96/dataset-pruning-sarcasm-detection

Description: No description provided.

Project Overview

This repository focuses on fine-tuning models for sarcasm detection using a pruned dataset.

Training Methods:

• The repository includes several pre-trained models:
  • RoBERTa
  • XLNet
  • Electra
• The models can be executed by running:

  python ./Models/model-name/model-name.py
  

• Dependencies are installed via:

  pip install -r requirements.txt
  

• The dataset used for training is SARC (Sarcastic Comments on Reddit), which can be downloaded from:
  • Kaggle Dataset.

Results:

• The best results from the paper can be reproduced by running:

  python ./Models/RoBERTa/RoBERTa.py
  

• The repository is based on a fork of SemEval 2022 Task 6 - Sarcasm Detection (GitHub).
• More details on the approach can be found in their related paper:
  SemEval-2022 Task 6 Paper.

Dataset:

• Source: Kaggle (SARC dataset).
• File: train-balanced-sarcasm.csv (not included in the repository but can be downloaded separately).