Compassion International is a non-profit organization, which connects an impoverished child with a sponsor that will stay with from early school age all the way until they are 18. Human translated letters between sponsors are kids have been the a key part of the 1:1 relationship we help children and sponsors foster.
Compassion International needs a solution which is more scalable and more immediate, while still protecting our children from any potential harm. Currently, this is typically done by manually reviewing all letters. They would like to add a voice strategy for connecting children to their sponsors, with near-real time translation from device to device, while also filtering out anything inappropriate for children to experience.
What it does
Walkie-Talkie will accept either text or voice recordings, translate them from the sender's language to the recipient's and scan it for context and malicious keywords.
How we built it
-Use Transcribe to turn speech to text -Use Comprehend to scan transcript for malicious intent -Use Translate to convert the message to the recipient's language -Use Wordpress for the frontend -Store translated files in S3
Challenges we ran into
-Time -Limited knowledge of AWS Translate, Comprehend and Transcribe (We wanted to learn!) -Issues with triggering Lambda's from S3 buckets
Accomplishments that we're proud of
-We made it to the end! -Walkie-Talkie can convert speech to text using transcribe
What we learned
-How to use AWS Translate, Transcribe and Comprehend -More about limitations of lamdba triggers for S3
What's next for Walkie-Talkie
- Convert text to speech so the recipient can hear the message, rather than read it, for a more personalized experience
- Increase localization
- Integrate with Amazon Echo