Inspiration
Instagram is easily a great platform for entertainment accessible to many young people, but there are many inappropriate or vulgar comments on Instagram Reels. Furthermore, these comments tend to gain many likes, so we often see them at the top of comment sections. Thus, we wanted to develop a way to hide negative comments and only show those that didn't contain hate speech.
What it does
Scroll Safe filters out comments with profanity to create a safe environment for watching reels. The website also has a cleaner user interface to get rid of potential distractions.
How we built it
We scraped data from Instagram using Instagram Bulk Scraper and Instagram Scraper 2022 from RapidAPI. We then Leveraged BERT, an advanced language processing model. After that, we fine-tuned the parameters on sample datasets from the web to fit our needs in order to have the best results. Lastly, we tested the AI on the Instagram data and made final adjustments.
Challenges we ran into
Some challenges we ran into were imbalanced datasets when training the AI model, sometimes communication with others, API rate limits when scraping data from Instagram, and lastly difficulties integrating AI when the files were too big.
Accomplishments that we're proud of
The two biggest accomplishments for us were that we were able to successfully filter out comments unsuitable for younger people and build a working AI model and frontend.
What we learned
We learned how to use APIs from RapidAPI, utilize BERT, and train an AI model with limited datasets.
What's next for Scroll Safe
We can train the AI model to be more accurate and consistent, generate reels tailored to the user’s interests, expand it to more platforms like YouTube Shorts and TikTok, and additionally filter inappropriate videos to make the content even safer.
Log in or sign up for Devpost to join the conversation.