Skip to content

Latest commit

 

History

History
59 lines (46 loc) · 3.64 KB

README.md

File metadata and controls

59 lines (46 loc) · 3.64 KB
Raw

Youtube Digest

A RESTful service designed to extract insightful data from YouTube videos. With just a YouTube video link as input, this API delivers multiple capabilities to help users better understand video content and audience sentiment. It is a powerful tool for content creators, marketers, and researchers looking to quickly gather actionable insights from YouTube videos and their engagement metrics.

Key Features

Video Content Blogging

The API analyzes the content of the video and generates a concise blog article summarizing its main points. This allows users to quickly grasp the video's topic without watching it in full.

Comment Sentiment Analysis

The API also performs sentiment analysis on the comments section of the video. It categorizes comments into three types:

  • Positive: Comments expressing appreciation or favorable opinions about the video.
  • Negative: Comments that reflect criticism or dissatisfaction.
  • Neutral: Comments that neither praise nor criticize but offer objective observations or questions.

Comment Summary

In addition to categorizing the comments, the API provides an aggregated summary of the overall sentiment, helping users understand the general perception of the video by its audience.

Workflow

Video Content Blogging

Black White Minimalist Boho Grid Background Page Border A4 (Landscape)

Sentimental Analysis on Comments

Copy of Black White Minimalist Boho Grid Background Page Border A4 (Landscape)

About the project

The API is built using Django Rest Framework and leverages multiple external services to accomplish the tasks described above.

Key External APIs and Libraries

YouTube Data API (googleapiclient.discovery.build)

  • Used to extract video details (e.g., title, description, and comments) and to fetch captions if available.
  • Retrieves up to 600 top-level comments for sentiment analysis.

AssemblyAI

  • A transcription service that processes audio extracted from videos to generate accurate transcripts.

MoviePy (moviepy.editor.VideoFileClip)

  • Extracts audio from downloaded YouTube videos for transcription purposes.

VaderSentiment (vaderSentiment.vaderSentiment.SentimentIntensityAnalyzer)

  • Conducts sentiment analysis on YouTube comments, categorizing them as positive, negative, or neutral.

Groq

  • An AI service for generating the blog article and summarizing the comments by processing video transcriptions and comment sentiment analysis results.

BeautifulSoup (bs4)

  • Cleans up the generated content (e.g., blog or comments summary) by removing any unnecessary HTML formatting.

Main Functions

get_result(url, title):

  • Main task that ties all processes together, generating the video transcription, summarizing the transcription into a blog article, performing sentiment analysis on comments, and summarizing the categorized comments.

generate_transcription(video_url, use_whisper):

  • Downloads the video, extracts the audio, and uses AssembyAI to transcribe the audio into text.

get_sentiment_analysis(link):

  • Retrieves YouTube comments, filters out irrelevant ones, and performs sentiment analysis using VaderSentiment, categorizing them as positive, negative, or neutral.

summarize_comments(positive_comments, negative_comments, neutral_comments):

  • Summarizes the comments into a general consensus using Groq’s AI service.

get_summary(title, transcription):

  • Uses Groq to generate a concise blog article summarizing the transcription of the video.