Introducing Pacify, a program aimed towards detecting and removing online hate speech.
Our Pacify project draws from the issue of hate speech over the Internet. In today’s modern world, the amount of online hate speech has increased dramatically, especially on social media or gaming platforms. Our Pacify project aims to reduce the negative effects of hate speech by detecting and censoring hate speech using a machine learning model and web extension.
Our web extension utilizes JavaScript, HTML, and CSS, while our machine learning model primarily uses Python. Separately, the machine learning model can predict whether a phrase is hate speech and the web extension can replace phrases at the click of a button. The web extension was built with JavaScript along with HTML and CSS for the user-side part of the web extension. For the machine learning algorithm, we first took a series of datasets and compiled them into one CSV file. Then, we used an algorithm to preprocess and normalize the data. We used the SVM algorithm to train the model. We saved all of this onto a pickle file for use later. Eventually, we plan on incorporating our machine learning component into our extension, though currently they work as separate programs. For both portions of Pacify, we referenced several online sources for inspiration and debugging.
While coding the web extension, we ran into several issues regarding the harmonization of our JavaScript and HTML files. Though these issues took us quite a while to solve, we eventually found a solution. However, another issue that took us a lengthy amount of time to solve was the machine learning algorithm. With no one in the group having much previous ML experience, it was difficult to get an algorithm started. Nevertheless, after lots of research, one of our members managed to code the algorithm. Ultimately, in the end, we weren’t able to achieve the product we had originally planned, but we managed to create two impressive products that even worked well separately. Through long hours of hard work, we eventually were able to code a working web extension and separate machine learning model. We all learned a lot from this experience, including but not limited to JavaScript, ML and AI.