CommsItumo
A web app for analysing sentiment and recurring topics in YouTube comments.
CommsItumo retrieves comments through the YouTube Data API. A Transformers model classifies them as positive, negative or neutral and extracts recurring topics from the comments.
After the analysis, users can ask questions about the results through a chat interface.
Technical details
I built the backend with FastAPI. Since an analysis can take some time, the work runs in the background and progress is sent to the client over WebSocket.
I added a three-level cache to reduce repeated processing of the same video. For repeated requests, this reduced processing time by roughly 40–50%.
The front end uses React, TypeScript and Vite. The application supports both Turkish and English.