Lingator

Making important documents easier to understand, act on, and access for everyone, regardless of language.

Comment

About

Lingator is a translation service designed to help navigate legal and medical systems. Our goal is to help users have a clear understanding of their rights, healthcare, and options available to them. When faced with challenges and hardships, many immigrants are unsure of the steps they should take and resources available to them as they are oftentimes unfamiliar with the Canadian law and their medical options. These uncertainties, often caused by language barriers and complex terminology, results in problems being left unsolved and immigrants feeling wronged. We strive to bridge this gap by providing accurate and intuitive translations and practical next steps, empowering users to make informed decisions.

Inspiration

We as a team bonded over our shared experience of being an immigrant or growing up as a second generation immigrant. From a young age, many of us had to step up for our parents when they needed to sign legal documents, translate at a doctor's appointments, and navigate systems that were never designed with them in mind. Understanding complicated concepts did not always come easily, not because of a lack of intelligence but because of language and cultural barriers. Knowing that not all immigrant families coming in have a reliable helper to assist them, which makes them vulnerable to being taken advantage of. Having been put in a completely different environment and lifestyle is overwhelming, especially when you don't understand the world around you. Thinking about our parents having to face this is really what motivated us to build something that empowers and protects them as well as many other families in this situation.

How We Built It

The backend of Lingator was attempted to be built using Next.js, a fullstack framework. Within Next.js, we used React, TypeScript, and CSS. Through TypeScript, Google Cloud API was connected and used to read documents, summarize (Gemini), and translate into the user language (Google Translate). Program generation was largely done using Claude.ai, and debugging efforts were aided with Github Copilot.

Challenges We Ran Into

Though we are first time hackers, we ambitiously hypothesized many ideas starting out this project that revolved around detecting user language based on a speech recognition model, acting as a service to immigrants and non-native English speakers. Our first iteration of Lingator was a desktop application created using Electron.js, but ended up being tossed aside after hours of trying to work with it because of lack of experience with the framework on top of fear of wasting time trying to learn. For the second iteration, we pivoted to having a web application instead built using Python, JavaScript, HTML, and CSS. That version was also discarded because learning to implement online APIs with Python communicating with JS was a difficult concept to grasp as well. With this being our first time using APIs in a project, we greatly struggled with coming up with a way to approach our problem statement. Even though it was difficult, the problem resonated with all of us, so much so that we wanted to persevere regardless.

For the final iteration of our project, we settled with using Next.js as a framework after researching and consulting with mentors, volunteers, and AI models. It seemed simple enough in theory and the directory could be generated into VSC with a couple of commands, so we moved forward. The biggest challenge we faced was at this stage. The learning curve was steep, from learning React to TypeScript to navigating the framework itself. After several hours back and forth with different AI models, we came across a barrier that refused to be solved. Specifically, the API was expecting the user input as a JSON file, but users could only input PDF files. The same error kept occurring no matter what was changed, with even more errors popping up on every attempt to debug. After 12+ hours trying to overcome the same error, we figured that it would be best to stop editing the program overall. To overcome this, we created a mock application HTML/CSS that clearly displays the interface and impact of Lingator.

What We Learned

Through the creation of Lingator, we as a team have learned not only hands-on practical skills but also resilience and to be more socially conscious. Through the workshops hosted and our own trial and error during the creation process, we improved our coding and programming abilities, specifically VScode, AI usage, frontend and backend development, and AI integration. Most of which are skills that have only developed in the duration of the competition.

As we faced technical and time-related challenges with the development, we had to pivot and change our original idea. This experience taught us how to think on the spot, when it is time of the essence. To not be focused on one idea but rather have a broader view on how to approach the problem, thinking outside the box. Having the courage to ask for help when you are stuck at a wall and need an outside view on things.

Finally, the most meaningful lesson we learned is about the struggles caused by language barriers. Building Lingator deepened our understanding of the challenges and struggles those in Canada with English as a second language face on a daily basis, reinforcing the importance of accessible technology for every person.

What's Next For Lingator

In the future, we plan to expand Lingators database to support many more languages to be more accessible for a wider range of users. We also hope to implement a speech-to-text recognition system that allows for users to speak in their native language while the AI automatically detects and translates it in real time. Once we can get users to test our product in Canada, we hope to expand it further beyond our borders. Our vision is for the platform to be usable worldwide and accessible to anyone, anywhere around the world.

Built With

Share this project:

Updates