Assistant Intent Classifier

Speech intent classificiation

Code Structure

├── Makefile
    - Common useful commands
├── app
    - Contains server code for Web interface 
├── notebooks
- Jupyter notebooks used to create/train the models
├── server.py
-  Run the Web Interface server locally
├── skills
- Skills created such as shopping and time
├── transcription
- transcription via deepspeech 

Local Setup Instructions

Install the requirements via pip3 install -r requirements.txt and run the jupyter notebook

The code can either be run as a live server mode, streaming mode, or direct text mode.

and run server via

pip3 server.py

or a local pipeline via

python3 text_based_classification.py

To get the models required to run, download the two model files at https://drive.google.com/drive/folders/16glf6J5UXeyCPusQR2xr8GGrD0r1hvXT and insert them into the skills folder.

• TODO simplify this process

The audio files that need to be processed should be placed inside the transcription/ folder currently(to support volume mounts within docker).

• TODO this will be updated later to allow for any path.

Architecture Documentation

One Architecture Diagram: https://people.eecs.berkeley.edu/~nmalkin/alva/architecture

Another Design/Diagram