-
-
The chat - Up and running!
-
Landing
-
The Problem
-
Features (Sorry it's so small, couldn't get it all on one image)
-
Built With + Future
Inspiration
According to UNESCO's 2020 report, an estimated 773 million adults worldwide are illiterate, which represents about 16% of the global adult population. The literacy rate is also closely linked to economic development and poverty. Low-income countries have lower literacy rates than high-income countries, and there is a strong correlation between literacy rates and levels of economic development. Despite this, most services that teach languages to adults are priced at standpoints that the people we are trying to help cannot afford.
Addressing the issue of illiteracy requires a multi-faceted approach that involves improving access to education, providing support and resources for adult learners at a price that they can afford, and promoting awareness and advocacy around the importance of literacy skills.
What it does
Chat: Linguix can communicate with you! Ask it a funny question, ask for a reccomendation, and try to keep the conversation going while observing the way it speaks.
Translate: Translate english text into one of the languages that Linguix has been trained on, including but not limited to, Spanish, French, German, Italian, Portuguese, and Chinese (Simplified and Traditional).
Improve: Linguix can improve text based upon grammar, spelling, style, word choice, tone, or how truthful it judges the text to be.
Generate: Generate well-written responses based upon simple text prompts in order to get an understanding of common framing and tenses used in the English language.
Simplify: Linguix can paraphrase long texts and remove larger, more complicated words to make literary resources more usable.
Encouragement: Get language-based compliments encouraging you to practice and never give up on trying to improve your English.
How we built it
Using a .JSON file, I tokenized the words and extracted features to create a neural net with 2 hidden layers that I implemented in Pytorch and trained into a .pth file. I also used OpenAI's Text Completion API and trained it for answering questions that my NLP couldn't. I made a function using the AI technoligies for both of these that was then implemented into the Flask framework. The HTML and CSS of the site displays a button - once clicked, it calls JavaScript which POSTs data to the python file that runs the AI function and returns the data to the JavaScript, which displays it using the HTML all over again.
Challenges we ran into
This was the first time I used JavaScript for interactions with Python, and I found it quite difficult to pass data back and forth between the technologies.
Accomplishments that we're proud of
I created a NLP!!! Which was very cool, something I have never done without a pretrained model before.
What's next for Linguix
In the future, I would want to improve my neuralnetwork to include more phrases for better, more relevant outputs. Additionally, I would try to see how I can improve upon OpenAI's technology by filterig out uneccessary information. Finally, I would like to add a login feature so that users can see how long they have been practicing and how their chat history has demonstrated their improvement.
Log in or sign up for Devpost to join the conversation.