AI assited vocubolary learning app.

Inspiration

I needed to learn vocabulary for GRE exam preparation. At the school level, I learned vocabulary using different techniques such as fill-in-the-blanks and making sentences with words. Word games were also very helpful in improving spelling mistakes. So, I planned to bring all of these features into one app.

Also, there are many websites that provide important word lists, and users sometimes get confused about which one to follow. That is why I decided to bring all the important word lists into our app.

What it does

Users need to log in by creating an account. After that, the app loads all vocabulary lists, including Barron 300 words, Magoosh 350 important words, and other suggested word lists. These are shown with labels in the vocabulary list tab.

By clicking on a word, users go to the word details page, where they can see the definition, example sentence, synonyms, and antonyms. Users can also write a sentence using the word, and AI evaluates how correct that sentence is.

After that, users can add the word to the learning queue. They can add a maximum of 5 words in the queue, so they can check what is inside the queue at any time. From the learning queue, users can select a minimum of 3 words and a maximum of 5 words to transfer to the testing zone.

In the testing zone, using the selected words, the AI agent generates stories, fill-in-the-blank exercises, and paragraphs using the selected words. The AI also evaluates the paragraph written by the user. There is also a mystery zone, where users need to find a word generated by the AI, along with word games.

After completing the tests, the selected words are transferred to the revision zone. From the revision zone, users can again go to the testing zone for revision. In every tab, users can see how many words they have learned.

How we built it

First, I collected all the words from GitHub and other free resources. Then I wrote a Python program to collect synonyms and antonyms from WordNet, so all synonyms and antonyms could be added automatically.

After that, I generated definitions for each word, mostly from WordNet, and example sentences using the Gemma 3 model.

Then I designed both the backend and the frontend. At first, in the testing zone, I only added fill-in-the-blanks, story generation, and AI feedback. Later, I added games and the mystery zone.

First, I tested everything using a local model. Everything worked properly, and I used it for personal learning. Later, I planned to publish it on the Google Play Store, and I uploaded the code to Google Cloud Console.

Challenges we ran into

I faced challenges in UI selection, backend deployment, and choosing which AI model to use. Although we deployed the code to Google Cloud Console, my free credits were almost finished.

I also faced integration challenges in loading vocabulary from the backend to the frontend, especially while integrating Firebase authentication.

Accomplishments that we're proud of

I am proud that I made this app, which is really helpful for me. Four of my friends are already using this APK version, and they are very impressed.

What I Learned

• Full-stack application development
  
• Using AI models for NLP tasks
  
• Working with WordNet and dataset processing
  
• Cloud deployment using Google Cloud Platform
  
• Mobile app publishing workflow
  
• Secure authentication and system integration
  

What’s Next

In the future, I plan to add:

• AI-powered voice pronunciation evaluation
  
• Real-time speaking practice
  
• Smarter adaptive learning system
  
• Better personalization for users
  

Built With

• docker
• fastapi
• gcs
• langchain
• langgraph
• postgresql
• python
• react-native