gianlucagiudice / irony-detection

Irony Detection

More info about the project can be found here: Thesis

Requirements

• python3.7

• Java 6 or above

• Libraries listed in requirements.txt

  Run pip3 install -r requirements.txt to install all python libraries

Input

Put the dataset into:

data/raw/DATASET_NAME/

Also write the associated labels in the _lables.json file within the same folder.

Output

data/processed/DATASET_NAME/

How to execute the program

Single script

Run run.sh. This involves:

1. Feature extraction using three different strategy for text representation:
   1. BOW
   2. BERT
   3. Sentence-BERT
2. Training of the models
3. Perform PCA

Manual version

• Feature extraction

  /main.py TARGET_DATASET TEXT_REPRESENTATION
  

  Parameters:

  • TARGET_DATASET = The name of the dataset

  • TEXT_REPRESENTATION = Strategy used for text representation. Valid values are:

    • bow
    • bert
    • sbert

• Training

  /training.py TARGET_DATASET
  

• Weka experiment converter

  /notebooks/weka_experiment_converter.ipynb 
  

  This notebook is used to convert .csv format of the weka experiments to the .json format which is consistent with the output produced by the scikit-learn reports output.

  In order to convert the reports just place .csv weka output into the folder reports/TARGET_DATASET/weka_experiment.csv.

  Be aware that the name must be weka_experiment.csv

• PCA

  /pca.py 
  

Analysis

In order to analyze the reports and PCA output, several notebooks have been created.

• Models report - Performance comparison

  /notebooks/report_analysis.ipynb
  

• PCA

  /notebooks/pca*.ipynb