Asmaai is a simple 2-layer neural network that generates Egyptian first names.
Before running, the following dependencies need to be satisfied:
- python3 needs to be installed on the environment you will run the script in (local, cloud, etc.) via command
python3.
Assuming python3 is installed via command python3, navigate to current directory and run:
source .env/bin/activate
python3 script.py
to run script in a virtual environment. Please note that the script takes up time and RAM - running on a cloud environment like Google Colab with a GPU runtime would be helpful.
When done, run source .env/bin/deactivate to deactivate virtual environment.
The file names-present-big.txt represents the dataset used for training, development, and test. To regenerate it:
- Download the name dataset from https://github.com/philipperemy/name-dataset.
- Unzip the folder.
- Move the
EG.csvfile to the current directory.
Asmaai is a portmanteau of the Arabic word "أسماء" ("asmaa" or names) and AI.
- Egyptian name data pulled from https://github.com/philipperemy/name-dataset.
- Code adapted from https://github.com/karpathy/makemore.
- Developed as part of the NLP course of the Universitat Pompeu Fabra's Master in Theoretical and Applied Linguistics taught by Professor Thomas Brochhagen.