SafeSpace

Inspiration

We were inspired by the heartbreaking reality that many survivors of domestic abuse suffer in silence due to fear, shame, or a lack of supportive, non-judgmental resources. Traditional reporting systems are often cold, complex, and one-sided. We wanted to create a tool that not only listens to survivors but understands them. SafeSpace was built to offer compassionate insights while flagging behavioral red flags—empowering survivors to make informed decisions.

What it does

SafeSpace is an AI-powered reporting and support platform for individuals experiencing abuse, manipulation, or harassment. When a user shares what happened, the platform uses Natural Language Processing to:

Detect emotional and psychological signals
Assess severity, coercive control, and abuse likelihood
Analyze sentiment and emotion match (e.g., fear vs. numbness)
Infer potential legal violations and contextual risk factors
Offer DSM-informed explanations and safety recommendations
Allow users to download a PDF report for documentation or legal use All insights are presented in a calming, supportive UI with empowering language and visualizations like severity bars, emotion badges, and context flags.

How we built it

Frontend: HTML/CSS with collapsible insights and responsive design using VSCode
Backend: Python with Flask
AI Models: Emotion detection: j-hartmann/emotion-english-distilroberta-base Toxicity detection: unitary/toxic-bert Abuse classifier: SkolkovoInstitute/roberta_toxicity_classifier Sentiment: cardiffnlp/twitter-roberta-base-sentiment-latest
PDF Export: xhtml2pdf for safe record-keeping
Custom Heuristics: Added rule-based logic to boost detection accuracy of coercion, threats, and psychological abuse

Challenges we ran into

Getting model outputs to reflect real-life severity levels was tricky. Most outputs returned low scores regardless of input. We overcame this by combining raw model scores with custom heuristics that parse phrases like “threatened to kill,” “screamed for hours,” or “controls who I talk to.”

Accomplishments that we're proud of

We created a platform that feels human and not clinical.
Users can walk away with a tangible, AI-powered summary that’s easy to understand and save.

What we learned

Building something that has real-world utility in sensitive spaces demands continuous reflection and collaboration.

What's next for SafeSpace

Our mission is to turn SafeSpace into a scalable, lifesaving companion that survivors can trust when the world feels unsafe. Future additions that we plan to implement to work towards this mission are: