Comments stop being noise once they become a decision queue.
NLP · Data product
An end-to-end NLP experiment, from a synthetic dataset to an analysis dashboard and YouTube comment classification.
Show the full cycle of a data product without hiding the limits of a synthetic dataset and an experimental model.
The project combines data generation, preprocessing, training, model comparison, a seven-page dashboard and integration with the YouTube API.
Present the solution as a pipeline experiment, not a market-ready model.
Prioritize an interface that drives action, such as the critical comments queue.
Treat API failures and missing credentials as product states.
Read perfect metrics on synthetic data as a methodological warning.
Overview, channels, topics, queue, word clouds and YouTube
Labeled synthetic comments
Logistic regression and Naive Bayes