All Projects
NLPDeBERTaYouTubeMultilingual

DeBERTa Video Classifier

AI YouTube Video Remark Classification

A Streamlit NLP app that scrapes YouTube channels, classifies videos into 31 content remark categories using a fine-tuned DeBERTa-v3-base transformer, supports multilingual content (Hindi, Punjabi, Tamil, Telugu, Marathi, and more) via auto-translation, and allows users to retrain the model with new labeled data — all from a single UI.

Type
AI / ML Platform
Category
AI & Machine Learning
Year
2025
Stack
Python · Streamlit
DeBERTa Video Classifier
At a Glance
31
Categories
DeBERTa
v3-base model
8
Languages supported
Live
Model retraining
What's Built

Key features & capabilities.

01

Fine-tuned microsoft/deberta-v3-base on Title + Description + Tags (512 tokens)

02

Auto language detection + translation for Hindi, Punjabi, Tamil, Telugu, Marathi, Kannada, Bengali

03

4-page Streamlit app: Scraper → Prediction → Retraining → Full Pipeline

04

Model versioning — retraining never overwrites previous model versions

05

Confidence scores and Top 3 predictions per video

06

Full scrape + predict pipeline: channel URL → classified Excel in one click

Want something like this?

We build full products — from idea to deployment. Let's talk.

Ready to build?

Ready to build what's next?

Tell us about your product. We'll come back with a plan, a team and a timeline.