Haven

Helping Domestic Violence Victims through ML.

Comment

Inspiration

Our team is made up of a 7th grader, two 8th graders, and a 9th grader. We had limited experience in ML, so this project challenged our abilities. {We spent a large amount of effort building the location suggested. Through numerous YouTube tutorials and StackOverFlow posts, that aspect was finally functional.} Sam H. Lawson Middle School Miller Middle School Dartmouth Middle School American High School What is your Discord username(s)?

  • SwarnyaSrivastava
  • Anishag01
  • happygurl66

What it does

It allows victims to go through and find safe places, find the cause of abuse and more. According to the CDC, 1 in 4 women and 1 in 7 men will experience physical violence by their intimate partner at some point during their lifetimes. Domestic violence is a serious problem in our world. Fear you are a domestic violence victim? Don't worry, Haven can help. Read on to find out how YOU can get help using our website in three easy steps.

  • The text analyzer takes a text and figures out if you are being domestically abused. This is especially helpful online if you were talking to someone and didn't know whether or not what they said was appropriate, you can copy-paste the text and press the analyze button to get your results.
  • On this page, you will find a survey that will give you personalized results after you are done. It is completely confidential, and you can get results right there and then.
  • Here, you can search for city facilities in your area. Currently, this only works for new york. Just search for your borough and you'll get a list of them.

How we built it

We built our website from scratch. We followed a tutorial on Natural Language Processing for the machine learning portion. We used Kaggle’s sentiment datasets to train our algorithm, resulting in an algorithm with an 83% accuracy.

Challenges we ran into

{We had a tough time designing the algorithm that directs the user to the closest help centers. We {solved}{tried to solve} this through countless hours of grit (and neglect of sleep).} Our team had very minimal knowledge of ML as we were almost all middle schoolers. At first, we had low accuracy, but after tripling the dataset, pipelining the process, and adjusting parameters we finally achieved an 83% accuracy. We spent numerous hours trying to use Django, Flask, and Jinja to link our Python ML code into the website. None of us had ever used it before, and it looked rather intimidating. We still attempted it, all the way until the submission deadline.

What's next for Haven

In the future, we’d like to partner up with an anti-domestic-violence organization. We also hope to collect more domestic violence data so that we can further recognize the type of abuse. Our main hope is to create an app that recognizes abusive speech.

Share this project:

Updates