Skip to content

Hack TCNJ '17. Neural network Alexa skill that detects harassment and reports it if detected.

Notifications You must be signed in to change notification settings

hdubel94/Harassment-Monitor

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

47 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Harassment Monitor

Hack Harassment Monitor

Inspiration

In a rapidly evolving digital world, it can become easy to forget that there is another individual on the other side of the screen. This can lead to the victimization of individuals through online harassment. As the world becomes ever more connected it has become of even greater importance to catch, report, and prevent any form of online harassment, ensuring that each individual can have a safe and enjoyable experience.

What it does

Our hack uses a neural network which is trained to identify potential harassment by utilizing static semantic analysis. To help demonstrate the network, it was implemented onto an Amazon Echo. Upon using the Echo's wake word and asking for "Harassment Monitor", the Echo will listen silently in the background for any words spoken and forward them into the neural network. If the network deems that the conversation could be considered harassment, the user is notified by Alexa and an e-mail and an SMS text message is sent to the supplied contact. We hope that by catching harassment and reporting it, the spread of it can be prevented for the greater good of society.

Hack Harassment Monitor Message

How we built it

We used python to implement our entire project. The neural network is designed in a separate file so that it can be used in myriad of applications. The HarassmentMonitor.py script is what AWS Lambda utilizes to collect and send information to the network and send any messages should they be needed.

Challenges we ran into

One large challenge is that Amazon's Alexa service cannot listen silently in the background indefinitely. This is to prevent a breach of individual privacy. Since our Alexa Skill cannot be left running, if Alexa does not hear any input our skill will be exited.

Another challenge was training the network. Many words can have both a negative and a positive connotation. This made it difficult at times for the network to properly determine whether what is being said is harassment.

Graph of Convergence

Accomplishments that we're proud of

This is the team's first implementation of the Alexa Voice Service and utilizing some of the features AWS has to offer. Successfully designed and created a neural network to classify harassment speech despite the inherent difficulties of extracting meaning from text.

What we learned

Gained further familiarity with Python. Learned the basics of an Alexa Skill and how to debug it. How to implement AWS features into our project.

What's next for Harassment Monitor

The neural network will be expanded upon and trained. Ideally, it could be utilized through other voice and text recognition scenarios and help fight harassment across multiple mediums.

The Harassment Monitor team absolutely repudiates all form of harassment. Our goal is to raise awareness and stop potential harassment in any setting. Any negative language or phrases used or found in the source code is purely for classifying or training data purposes. The Harassment Monitor team does not approve of any use of such negative or harassing language in any form.

About

Hack TCNJ '17. Neural network Alexa skill that detects harassment and reports it if detected.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published