YouTube Sentiment Analyser

---

https://youtube-sentiment-analyser.thankfulcoast-599cc73b.westus2.azurecontainerapps.io

---

YouTube Sentiment Analyser is a web application that analyzes the sentiment of comments on YouTube videos. Users provide a YouTube video URL, and the system processes the comments using pre-trained NLP models to classify sentiment and detect trends.

Usage Example

To analyze the sentiment of YouTube video comments, provide a YouTube Video URL, wait for the system fetch comments from the video, models classify the comments, and the system displays the sentiment trends and visualizations.

Process of Sentiment Analysis

Results

Key Features

• Sentiment Analysis: Classifies comments as positive, neutral, or negative.
• Spam Detection: Filters out spam comments.
• Hate Speech Detection: Identifies and adjusts for hate speech and offensive content.
• Sarcasm Detection: Modifies sentiment predictions for sarcastic comments.
• Multilingual Support: Analyzes non-English comments separately.
• Visualization: Displays sentiment trends for insights into audience reactions.

How It Works

The system uses five models that work together to classify YouTube comments. These models detect spam, hate speech, sarcasm, and sentiment, applying a weighted approach to determine the final classification.

For a detailed explanation of how the models make predictions, visit the models folder, which contains a full detailed explanation of the decision process.

How to Download the Docker Image

To run the YouTube Sentiment Analyser using Docker, you can choose between two versions:

1. Faster GPU Version (Requires CUDA 11.8)

For users with a compatible GPU and CUDA 11.8, this version runs faster:

docker pull davydantoniuk/youtube-sentiment-analyser:cuda11.8

2. Lightweight CPU Version (No GPU Required)

For systems without a GPU, this version runs on the CPU:

docker pull davydantoniuk/youtube-sentiment-analyser:cpu

More details, including instructions on setting up an API key, can be found on Docker Hub repository.