Ziad236 / arabic_dialects

Arabic Dialect Classification

Arabic tweets dialect

Many countries speak Arabic; however, each country has its own dialect, the aim of this Project is to build a model that predicts the dialect given the text.

✨Dataset✨

addressed by Qatar Computing Research Institute, moreover, they published a paper, feel free to get more insights from it,

https://arxiv.org/pdf/2005.06557.pdf

Repo Structure

• Data: This folder contains the data.db used for the project.

• Models: This folder contains trained machine learning / deep learning models used in the project.

• templates: This folder contains HTML templates used for rendering web pages in the application.

• Data_Fetching.py: This file contains code for fetching data from external sources.

• NLP_Final.ipynb: This file is a Jupyter Notebook containing the final version of the NLP (natural language processing) code used in the project analysis.

• app.py: This file is the main script for running the web application.

• data_cleaning.py: This file contains code for cleaning and preprocessing the raw data before it is used in the models.

• dl_model.py: This file contains code for defining and training a deep learning model for the project.

• helper_functions.py: This file contains helper functions used throughout the project.

• ml_model.py: This file contains code for defining and training a traditional machine learning model for the project.

• requirements.txt: This file contains a list of all the dependencies required to run the project.

Installation

To install the dependencies listed in a requirements.txt file, you can use the pip package manager in the command line. Here are the steps:

1. Open a terminal or command prompt.
2. Navigate to the directory where the requirements.txt file is located.
3. Run the following command to install the dependencies:

pip install -r requirements.txt

For run :

python3 scr/app.py

Models Comparison

We tried:-

2 ML Models

• Naive bayern
• Linear SVC

2 Dl Models

• Vanilla RNN
• LSTM

Model	F1_macro_val	accuracy_val
Naive bayern	51	67
Linear SVC	80	84
Vanilla RNN	50	52
LSTM	75	80

Team:-

1. Amr Mohamed Abd ALbadee
2. Amira Hesham Mo’men
3. Ibrahim Ayman Abu-Shara
4. Mostafa Mohammed ali
5. Ziad Mahmoud Mohammed