Inspiration
Every day, millions of people receive emails with an aggressive, ambiguous, or manipulative tone. These messages cause stress, misunderstandings, and sometimes workplace conflicts. CleanMail was inspired by observing:
the difficulty students, professionals, and freelancers face in managing digital stress,
the lack of tools capable of analyzing emotion in emails,
the need for a simple, fast, and reassuring solution.
We wanted to create a tool that not only detects toxic messages but also helps users respond calmly and professionally.
What it does
CleanMail is an intelligent email analyzer that:
automatically detects negative emotions (anger, threats, pressure, manipulation),
calculates a toxicity score from 0 to 100,
highlights problematic phrases and explains why they are concerning,
generates polite and constructive response suggestions to defuse conflict,
classifies emails as “safe,” “urgent,” or “at-risk.”
In one second, users understand the true tone of the message and how to respond calmly.
How we built it
CleanMail was built using:
HTML/CSS/JavaScript for a simple, responsive, and intuitive interface,
Python (Flask) for the backend,
NLP (Natural Language Processing) with spaCy / scikit-learn to analyze emotions,
a classification model trained on a dataset of aggressive, neutral, and professional email examples,
a custom algorithm that generates context-aware response suggestions.
The frontend sends the email text to the Python API → the AI processes it → returns the analysis and suggestions.
Challenges we ran into
We faced several challenges:
Detecting sarcasm and subtle nuances, which are difficult even for humans.
Balancing emotion categories to reduce false positives.
Keeping the analysis fast, even for long emails.
Providing natural, non-robotic response suggestions.
Ensuring total privacy: no emails are stored.
These challenges led us to simplify, optimize, and test continuously.
Accomplishments that we're proud of
Creating a useful and powerful tool using only web + Python skills.
Achieving instant tone analysis, even on long emails.
Successfully generating automatic responses that feel authentic and polite.
Making CleanMail accessible to non-technical users.
Building an innovative project that helps reduce digital stress.
What we learned
Through this project, we learned:
how to train a simple but effective NLP model,
how to structure a Python API and connect it to a web interface,
how to detect different emotional tones in text,
how to turn a complex idea into a functional MVP,
the importance of UX: simplicity drives adoption.
What's next for CleanMail
In future versions, we aim to:
integrate CleanMail directly into Gmail / Outlook as an extension,
improve detection of sarcasm and subtle manipulations,
add a coaching mode to help users write better themselves,
provide an emotional history of received emails,
offer response templates tailored to user personality,
launch a mobile version.
The ultimate goal: become the universal emotional filter that protects every user from digital stress.
Log in or sign up for Devpost to join the conversation.