• Automatic speech recognition for Albanian language with our best trained model
• User-friendly interface to label and validate speech datas
• Dataset with Albanian Speech datas.

This project is made up of three main parts: an API that serves all the files to the UI, a section that is used to label and validate Albanian speech data, and the third part which is the model test. In the web folder are all the files needed to run the UI, in the api folder are the files needed to run the faastAPI, and in the automation folder are some scripts used for many things, including generating datasets. The folder training includes all the notebooks used to train the model.

Docker is required to run the project locally. To run the project locally run:

docker-compose up --build -d

After all the containers are up and running you should find at:

localhost:140 -> API

localhost:120 -> WEB

The PRODUCTION WEB interface: dibraspeaks.uneduashqiperine.com

The PRODUCTION API: api.uneduashqiperine.com

https://www.kaggle.com/flooxperia/

We welcome all contributions to this project. If you have any ideas for new features or improvements, please feel free to create an branch, and test it by creating a pull request to merge your ticket to dev. we will try to approve the request as soon as possible and the change will be for some time in dev. After the feature or the improvement have been tested on dev than we can merge it to production.

finalyearproject.pdf

Our model Architecture VS Deepspeech on our dataset

• @florijanqosja

https://www.buymeacoffee.com/florijanqosja https://github.com/sponsors/florijanqosja

This project is licensed under the MIT License.