/BiBiNET

💬A Gated Recurrent Neural Network for Supervised Text Classification: detecting hate speech from different online textual genres.

Primary LanguagePythonMIT LicenseMIT

BiBiNET: BiLSTM for bipolarity sentiment analysis.

A Gated Recurrent Neural Network for Supervised Text Classification: detecting hate speech from different online textual genres.

License GitHub stars GitHub forks

Description

This project was developed for the ”Human Language Technologies” course of Professor Giuseppe Attardi.

Directory structure (main elements)

BiBiNET
  │── src
  │    │── data_import.py                   # process 1/6
  │    │── data_prep.py                     # process 2/6                     
  │    │── preproc.py                       # process 3/6
  │    │── transform.py                     # process 4/6
  │    │── classifiers.py                   # process 5/6
  │    │── test.py                          # process 6/6
  │    │── utilities.py
  │    └── main.py                          # file to run
  └── data
  │    └── forum_data
  │    │   │── all_files.csv                # text
  │    │   └── annotations_metadata.csv     # labels
  │    └── twitter_1
  │    │   └── twitter_dataset.csv       
  │    └── twitter_2
  │    │    │── train.csv                
  │    │    └── test.csv     
  │    └── wikipedia_data
  │         │── train.csv                
  │         └── test.csv    
  └── glove
  │   │── glove.twitter.27B.100d      
  │   └── glove.twitter.27B.200d        
  └── model          
  │   └── model.h5                          # final model   
  └── requirements.txt
  └── report.pdf                            # project guide
  └── README.md
  └── LICENSE

Quick start

Install Python:
sudo apt install python3

Install pip:
sudo apt install --upgrade python3-pip

Install requirements:
python -m pip install --requirement requirements.txt

Execute main

cd src/
python main.py

Corresponding author

Dr. Diletta Goglia ORCID logo
Postgraduate Student in MSc in Artificial Intelligence
Computer Science department, University of Pisa, Italy
d.goglia@studenti.unipi.it
dilettagoglia.netlify.app