Inspiration: Our primary inspiration for our project revolve around the fact that there is a lot of mistyped information in the world. In fact, Medical prescription errors are the THIRD leading cause of death in America. Typos have cost the government BILLIONS of dollars over the past 2 decades. In order to remedy this, we crested TYPE. ai.
What it does: Basically, what TYPE.ai does is that it first gives a user a diagnostic quiz based on what topic the user chooses. For example, if the user chooses Biology, then a Biology diagnostic quiz will appear. These "quizzes" are essentially where the user has to type a certain passage that it given on the screen as fast as possible without making any errors. These passages are generated through AI and Natural Language Processing. We keep a track of what words a user is getting incorrect so that we can display them at the end. Given the particular key words that the user struggles on, another passage will be generated using NLP to improve spelling comprehension.
How we built it: We used several NLP models such as a large Facebook dataset and T5 to generate the texts. We also had to use a Dictionary.com API where we got a definition of a certain topic so that the NLP model has some context to how to generate the text. Also we analyzed the users keystrokes to perform analysis about which word the user was getting wrong.
Challenges we ran into
Backend - NLP
- Passage Generation: We had to find and use 3 different models to generate the paragraphs, while passing inputs through them.
- API Usage: We used the dictionary.com API in lieu of a module that wouldn't install on our computers Backend - Word Selection
- Keyboard input: mscvrt
- Worst Word Selection: We had to use modules for this too, however, they were all from python natives, so we had to very thoroughly answer each question
Accomplishments that we're proud of: We were very happy that we were able to finish everything on the backend and its performance overall.
What we learned: We learned a lot about Natural Language Processing from this Hackathon
What's next for TYPE.ai
- Custom model creation for improved accuracy with specific inputs
- A login system for users
- Database storage to store user and company profiles
- Direct communication and integration with widely used synergy platforms
- A more cohesive and intuitive frontend
Log in or sign up for Devpost to join the conversation.