This is an implementation of a deep learning model for trigger word detection. The model enables, after training, to detect a trigger word mention in a speech. The model is currently trained using a digits dataset, and is being improved upon using precision and F1 scores criterias. Any improvements are very welcomed.
The model's precision and F1 score are being improved further.
Require Python 3.5+ and Jupyter notebook installed
The first cells of the notebook contain the installation scripts, as well as scripts for downloading the dataset necessary for training. They could be run in succession the take care of this phase.
In the project directory start a command line, then run command
jupyter notebook
Open the main notebook in this repository.