CogniCloud

Organizing mental clutter into neatly labeled clouds. An accessible tool to centralize one's ‘brain dumping’ and recognize patterns in thoughts, dreams and ideas.

Comment

Inspiration

Our inspiration came from the need to address a growing issue: many individuals, particularly stressed college students, struggle to find the time and tools to assess their mental health. As a result, mental clutter builds up, often leaving people overwhelmed. We recognized that without a structured way to reflect on our mental state, it’s difficult to manage or improve it. By tackling individual thoughts as distinct problems, we can positively influence overall mental well-being. Our tool, CogniCloud, aims to provide clarity by categorizing thoughts based on emotions, helping users track patterns in their mental processes. This visualization not only offers clarity but also serves as a valuable reference for individuals to share with therapists, creating a clearer path toward emotional health.

What it does

CogniCloud takes user input in the form of free-form thoughts and applies sentiment analysis to uncover the range of emotions connected to each entry. It uses a two-layer neural network: the first layer performs multi-label classification to detect all applicable emotions, while the second layer is a regression model that determines the intensity or weight of each identified emotion.

Based on this emotional breakdown, each thought is visually grouped into cloud-like categories, offering users a more intuitive and organized way to reflect on their inner experiences. This visual structure helps reduce mental clutter by transforming raw thoughts into a clearer emotional landscape.

How we built it

We used Next.js, Python, Django, Figma, Auth0, MongoDB, HTML, CSS, JavaScript, Kaggle, Hugging Face API, Tailwind, TypeScript, Transformers, TensorFlow, and PyTorch.

Challenges we ran into

One of the biggest challenges we ran into was finding a good dataset that could capture a wide range of real emotions. A lot of the ones we came across were either too small, too basic, or didn’t allow for thoughts to be tagged with more than one emotion. We wanted to reflect how complex feelings can be, so we needed multi-label data. On top of that, figuring out how to assign the right amount of weight to each emotion in a thought was tricky.

Accomplishments that we're proud of

We’re proud of building a working prototype in just 24 hours that brings together machine learning and a simple visual user experience. Our model used a two-layer neural network where the first layer handled multi-label classification to identify the different emotions in each thought, and the second layer used regression to assign intensity levels to those emotions. Even though two of our team members were new to coding, we all learned quickly and built something meaningful together. We were able to integrate sentiment analysis, connect everything to a database, and design an interface that helps users organize and reflect on their thoughts. Seeing the idea come to life was exciting, especially knowing it could help students manage mental overload.

What we learned

Over the past 24 hours, our team learned a lot about integrating pre-trained models and training custom ones from scratch. We gained hands-on experience combining physical design ideas with working UI and UX, and explored ways to use large language models for analyzing thoughts. We also got to work with a variety of tools and tech stacks, which helped us understand how everything fits together. Since two of our team members were just starting with coding, this was a big learning moment for us. It pushed us to work closely as a team, support each other, and grow fast together.

What's next for CogniCloud

We hope to integrate features that allow users to receive specific advice based on their mood and geographical location.

Built With

Share this project:

Updates