HEARSAY
The web is for
everyone
Giving everyone a voice to the web
Aditya Khowal && Bhavik Soni
Dyslexic
Friendly
font!
Our Story
From our humble beginnings working on the website for a UW RSO to the UWB AI hackathon
Students @ UW
Making Software Engineering more accesible
HearSay
Making the web
accesible
Why
Hearsay
46,440,000.
that’s the number U.S. adults have a disability affecting their participation in daily activities, including web use.
96.8% of homepages exhibit detectable WCAG 2 failures, with common issues including low-contrast text, missing alt text for images, and poorly labeled forms​
Web Content Accessibility Guidelines (WCAG) Level 2 compliance sites are expected to outperform their market competitors by 50%
Gartner, 2023
Technology
Python
- Versitle libraries
- Runs on a lot of things
Selenium
- Parse and interact with sites
- Can access JS rendered content
LM Studio
- Lets us test and run many LLM’s locally and for free and fine tune parameters
LLama 3
- Open source model by META
- great at interpreting code
- big context size (html is huge)
Eleven Labs
- Lifelike speech generation
- Can use 29 different languages
Coqui-AI TTS
- Free Text to speech API
- open source
Future Iterations
Fully local app and fully deployed app
The app can go in either direction, deployed on the cloud through a service like azure, or be a standalone app running fully locally.
UI beyond HTML
HTML is only the tip of the iceberg. UI is everywhere, the OS, apps and more all use different systems. Maybe using vision rather than code?
Advanced AI capabilities
Give the bot more context, personal information, recommendations, your calendar, etc. With greater context, the bot can do more without you always specifying
Code and
Demo
