Hana Mengistu

The goal of this project is to create a model that can understand documents that are paragraph format or table format and extract the text to be useful

This practicum proposal focuses on the design and development of a predictive modeling system to assess the virality of a thumbnail. Thumbnails act as the critical determinant for clicks and overall engagement. In this project, we dive deep to see if the type of thumbnail determines the success rate or, in this case, high interaction or view count. This project builds a pipeline that allows users to input a thumbnail and receive a success rate calculation. The system uses historical data alongside CLIP and manually selected features to develop an XGBoost model to identify patterns between thumbnail images.The project presents a functioning prototype that allows users to select a long-form or short-form thumbnail option and upload a thumbnail to generate a success rate for the input, as well as providing some improvement recommendations.