This project is a capstone project for the UC Berkeley Master of Information and Data Science (MIDS) program, an interdisciplinary, online graduate program designed to equip students with the skills and knowledge needed to excel in the field of data science. Combining rigorous coursework with real-world applications, the program covers a broad range of topics, including machine learning, data engineering, data visualization, and statistical analysis. The MIDS program emphasizes the ethical use of data and the importance of communicating complex findings to diverse audiences.

We are a dedicated team of four students working on RespectNet as our capstone project for the UC Berkeley Master of Information and Data Science (MIDS) program.

RespectNet is a project with a crucial mission to create an inclusive, safe space for school-aged children online. In today's digital age, we’re getting involved in online communities at younger ages, often before we’ve been taught how to interact with one another. This project is a call to action, emphasizing the importance of learning to interact online in a respectful and responsible manner.

Our aim is not only to promote kindness and inclusivity but also to educate students on why their comments require improvements. It’s one thing to make a blind edit to another’s words, but to truly have them change and be better going forward, it must be explained why a rewording has happened. This educational aspect of our project will enlighten students on the impact of their online interactions.

We’ve used a combination of intensely trained LLMs to get the best output possible - Jigsaw’s Perspective API, ChatGPTs Moderation API, and Cohere API. Combining these APIs allowed us to identify the toxicity of a comment through the Perspective API, remove hate speech commentary with the Moderation API, and generate a proper and improved rephrased comment using the Cohere API.

We’re excited to share this with a broader audience and see the impact we can have by extending the research already being completed on classifying toxic comments online. Now, we’re improving and educating students about why these toxic comments are wrong and hurtful.