VideoQuest - Video Search Engine

Overview

This project is a video search engine application with a graphical user interface (GUI) using MongoDB, Neo4j, and MySQL for data storage and retrieval. The application includes features such as video indexing, relationship management, user authentication, sentiment analysis for user comments, and personalized video recommendations.

Important Links

• Video Demonstration
• Saved Models

Project Directory Structure

Root Directory

Directory/File	Description
hate_speech_model.joblib	Joblib file for the hate speech detection model.
individual.json	JSON file containing individual video data.
main.py	Main script for the video search engine application.
preprocessed_data	Directory containing preprocessed video data in JSON format.
relations.json	JSON file containing relationship data.
sentiment_analysis_pipeline.joblib	Joblib file for the sentiment analysis model pipeline.
sentiment_analysis_training.py	Script for training the sentiment analysis model.
users.json	JSON file containing user data.
users_relation.json	JSON file containing user relationship data.
video_data	Directory containing raw video data.

website Directory

Directory/File	Description
api.py	Flask Blueprint for sentiment analysis API routes.
auth.py	Flask Blueprint for user authentication routes.
creators.py	Flask Blueprint for user analytics routes.
mongodb_models.py	Class for MongoDB interactions.
mysql_models.py	Data models for SQLAlchemy and MySQL.
neo4j_models.py	Class for Neo4j interactions and search recommendations.
preprocessing.py	Script for video data preprocessing.
static	Directory for static assets in the Flask application.
templates	Directory for HTML templates in the Flask application.
upload_youtube_videos.py	Class for uploading YouTube video data.
views.py	Flask Blueprint for general web application routes.

User Interface Preview

Beautiful Backend and Landing Page

Showcasing the aesthetically designed backend and landing page of the video search engine.

Sign Up and Login Pages

Highlighting the user-friendly sign-up and login pages for seamless authentication.

Home Screen with Search Bar

Illustrating the home screen featuring an intuitive search bar for easy video queries.

Video Page with Suggestions and Interactions

Providing a glimpse of the video page presenting the selected video, related suggestions, and interactive options like liking, subscribing, and adding comments.

Upload Videos and Sentiment Analysis

Showcasing the functionality allowing users to upload favorite YouTube videos and creators to check sentiment analysis for their videos.

Features

Multi-Database Integration

• MongoDB:

  • Indexing Video Files: Efficient indexing of video files for fast retrieval.
  • Search Functionality: MongoDB used for search based on common words and relevant keywords. Jaro similarity-based scoring system employed to rank search results.

• Neo4j:

  • Managing Relationships: Nodes for 'USER,' 'CHANNEL,' and 'VIDEO.' Relationships established based on commonalities such as subscribed channels, liked videos, and similar content.

• MySQL:

  • Relational Information Storage: Storing click-through data and relational information, critical for improving search results and personalized recommendations.

Graphical User Interface (GUI)

• Components:

  • Search Query Panel (SQP): Allows users to input video search queries.
  • Search Button (SB): Initiates the search based on the query.
  • Search Result Panel (SRP): Displays a list of relevant videos.
  • Current Video Panel (CVP): Shows detailed information about the selected video.

• Workflow:

  • Users input search queries in SQP and click SB to view relevant videos in SRP.
  • Clicking on a video in SRP updates CVP with video details and refreshes SRP with related videos from Neo4j.
  • Click-through information is stored in MySQL, capturing details about user interactions and preferences.

User Authentication and Customization

• Authentication:

  • User login, logout, and signup functionality.
  • Secure password hashing.

• Customization:

  • Initial video suggestions based on common words.
  • User behavior tracking for customization as users interact more.

Sentiment Analysis

• Sentiment analysis on user comments using a machine learning pipeline.
• Trained model utilized in the Flask API to provide sentiment scores for comments.

Web Scraping and External APIs

• Web scraping for extracting video data from YouTube.
• External APIs used for correcting English in search queries.

Analytics and Collaborative Filtering

• User analytics for processing and analyzing user comments using a pre-trained machine learning model for hate speech detection.
• Collaborative filtering using Neo4j, creating relationships based on user interactions.

Caching Mechanism

• Implemented a JSON-based caching mechanism to store user activity and metadata changes.
• Data, such as video metadata and user activity, cached in JSON files for faster retrieval and updating.
• Optimized data fetching from Neo4j, particularly when dealing with slow Python abstractions.

System and Database Dependencies

Dependency	Purpose
Python 3.x	Programming language for the project.
mongodb	Database for storing video information and enabling fast queries.
mysql	Database for managing user-related data, login credentials, and user behavior tracking.
neo4j	Database for storing relationships between users, channels, and videos for the recommendation engine.

Getting Started

1. Install dependencies: pip install -r requirements.txt.
2. Configure database connections in respective files.
3. Run the Flask application: python main.py.

Authors

• Akriti Gupta: Pre-final Year (B.Tech. Artificial Intelligence & Data Science)
• Sagnik Goswami: Pre-final Year (B.Tech. Artificial Intelligence & Data Science)
• Tanish Pagaria: Pre-final Year (B.Tech. Artificial Intelligence & Data Science)
  (IIT Jodhpur Undergraduates)