Inspiration
Children need to develop essential skills like reading and writing, but with remote schooling, it's becoming more difficult for them to learn. Young children aren't able to identify their own mistakes and consequently need some sort of personalized feedback. Unfortunately, parents are busy with their own work and it's much more difficult for teachers to help remotely.
What it does
Spellr is a web app that gives personalized spelling feedback. The child can upload an image of their written work. Using AI, Spellr converts the handwriting to text, runs through the text to find incorrectly spelled words, and generates suggestions of correct spelling. Then, the child can select the word that they intended, reinforcing the correct spelling and allowing them to learn without the help of a teacher.
How we built it
We used Flask for the backend implementation and Bootstrap, Semantic UI, HTML and CSS to create the front-end. The AI handwriting to text conversion used the Google Cloud Vision API and we used the Python Spellchecker and Textblob libraries to identify incorrectly spelled words and generate the spelling suggestions.
What's next for Spellr
After generating the spelling suggestions, we'd like to have these recommendations read out loud, as this will make it easier for the child to identify which word they actually intended to spell. Secondly, we intend to implement the GPT-2 language model, to make the learning experience more fun and enjoyable for the child. The child could write a couple sentences of a story, and using this as input, the GPT-2 model could generate a few more lines of the story. This could continue back and forth, essentially gamifying the learning experience and encouraging the child to keep improving their learning skills.
Log in or sign up for Devpost to join the conversation.