Difficulties in accessing healthcare are one of the leading issues faced by Canadian refugees today. The language barrier tends to be a prominent cause of this, especially when people first emigrate to the country and are generally only fluent in their mother tongue. Furthermore, the lack of common terminology for medical symptoms in most languages means that communicating to a medical professional is often an exercise in frustration. Additionally, this can lead to delays in getting the help that is needed, which increases the chances of complications arising.
What it does
Decipher aims to eradicate these barriers to conversation by providing a speech-first platform that bridges the gap in communication by easy translation at the tip of your fingers. By leveraging the power of the Google Cloud Platform, we are able to seamlessly translate speech across a wide variety of languages, allowing for a more natural flow of conversation.
How we built it
This project consists of a Python backend which queries neural network translation models hosted on the Google Cloud Platform. A Flask server listens for requests from the React JS frontend and routes them to the appropriate model.
Challenges we ran into
- Learning how to use GCP models.
- Getting the models to talk to React in a seamless manner.
Accomplishments that we're proud of
Getting the product to a stage where it can be used for real-time conversation. This model works on a end to end model, i.e, it takes in speech and returns translated speech. This removes many more barriers for people who are unable to read or are visually impaired. Our end-to-end service is what sets us apart from our competitors.
What we learned
- We got comfortable to using speech models.
- We also learnt how to send files over POST requests.
What's next for Decipher
- We want to leverage regional models to improve the accuracy of translations of some languages.
- Expanding the usability to different areas such as consumerism (shopping malls) , airports, etc.
- Adding more supported languages.