/XLM-RoBERTa-and-DPCNN

A Duet of XLM-RoBERTa and DPCNN for Sentiment Analysis

Primary LanguageJupyter Notebook

Project title: "A Duet of XLM-RoBERTa and DPCNN for Sentiment Analysis"

  1. Setup the environment:
    source setup/conda-create.sh
    source setup/conda-active.sh

  2. Install all the dependencies:
    pip install -r requirements.txt

**NOTE: You have change the path to the dataset, according to your

Dataset
Inside a /data/ directory:
1. You can find raw, augmented and pre-preprocessed SST-5 dataset.
2. Moreover, we provide dataset statistics inside /statistics/ folder.
3. In addition, classification reports, confusion matrices and training history are provided inside /history/ folder.

Data Pre-processing
Inside a src/data_prep directory:
1. Inside an /augmentation/ directory, you will find an implementation for the EDA augmentation technique.
2. Inside an /pre-processing/ directory, you will find an implementation for the data pre-processing and statistics.

Training XLMR and DPCNN
Inside a src/training directory:
1. You will find a code for XLMR fine-tuning and XLMR-DPCNN combination