Skip to content

code built by team GRO during the 2019 BCG Digital Ventures hackathon

License

Notifications You must be signed in to change notification settings

joeborrello/ellipsis

Repository files navigation

Ellipsis Logo

Ellipsis

Verbal auto-complete for monitoring neurodegeneration

Built by Team GRO at the 2019 BCG Digital Ventures hackathon


Overview

Usage

When at a loss for words, initiate Ellipsis and say the incomplete sentence. A word to complete the sentence will be suggested.

User Interface

Converts audio speech to text and sent to a RESTful API for processing. Displays the returned suggested word.

RESTful API

POST

Type: ‘string’ Input an incomplete sentence that will be autocompleted by Natural Language Processing with the GPT2 model. Returns the suggested word.


How to run

  • Run minimalRESTAPI.py to start autocomplete server

The minimalRESTAPI.py script starts the server, which listens for a POST from the iOS application

  • iOS application, when prompted, listens for spoken words, which are subsequently transcribed to text and posted to the autocomplete server

  • When transcribed text is posted, the python script uses NLP model to suggest the next word, which is then sent back to the iOS application

  • when the word is received by the iOS application it is spoken back to the user

  • Running the iOS app:

    • Tap listen, tap record, speak, when you pause it makes the API request

Note: Running the full, mobile version of this app requires that the iOS app is loaded onto a mobile device (or mobile device simulator)


How to run (without an iOS device)

  • To see how the autocompletion server runs without using an iOS device, simply start the server with python minimalRESTAPI.py and, in a separate terminal window execute curl -X POST -d *"YOUR SENTENCE FRAGMENT HERE"* http://157.245.196.85:5002/predict

How to run (from a docker image)


Dependencies

  • As best as possible, running pip install -r requirements.txt will install all required dependencies on a system

Team

Built by Team GRO at the 2019 BCG Digital Ventures hackathon Joe Borello: Algorithm developer Taylor Pullinger: Design and business Ralph Hertz: Design and business Victor Zhong: iOS developer Vincent Tu: Algorithm developer

About

code built by team GRO during the 2019 BCG Digital Ventures hackathon

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published