Irony Detection of Arabic Tweets

This project investigates irony detection in Arabic dialect comments on Twitter to enhance digital communication understanding.We will collect tweets from Arab users and manually annotate them to create a dataset. Using this dataset, we will train a neural network, bert model, and an arabert model aiming for an initial accuracy of over 78% in classifying tweets. The project aims to contribute to sentiment analysis by addressing challenges in multilingual contexts.

This project also represents my senior project for a BS in Computer science at Phoenicia University.

Installation

You need python 3.7 for the CNN Model in your environment

To set up this project, follow these steps:

1. Clone the repository:

   git clone https://github.com/jad-raad/Irony-Detection-of-Arabic-Tweets-

2. cd your-repository

3. python -m venv venv
.\venv\Scripts\activate

4. pip install -r requirements.txt

5. Add the arabic embedding file

Arabic Embedding file

The Arabic word embedding file mentioned in the config (wiki.ar.vec) is from the FastText pre-trained word vectors. You can download it from the official FastText website. Here's how you can get it:

1. Go to the FastText website's pre-trained word vectors page: https://fasttext.cc/docs/en/pretrained-vectors.html

2. Scroll down to the section "Pre-trained word vectors"

3. Look for the Arabic language (ar) in the list

4. Click on the "bin+text" link next to Arabic. This will download a file named cc.ar.300.bin.gz

5. Once downloaded, extract the .gz file. You should now have a file named cc.ar.300.bin

6. After extraction, you should have a file with the .vec extension, which you can rename to wiki.ar.vec

7. Place this file in the /data/fasttext/ directory of your project (or update the config file to point to wherever you place it)

All Bert Models Installation

It is recommended to run BERT models on google colab in order to utilize the T4 gpu of colab

1. Upload the entire project without the fasttext file and the environments into your google drive

2. Connect google colab to your personal drive

3. Run the allBertExamples.ipynb file and check the results

4. NOTE: google colab already supports all requirements for the bert examples, thus there is no need for installing libraries