Skip to content

Latest commit

 

History

History
45 lines (38 loc) · 1.72 KB

README.md

File metadata and controls

45 lines (38 loc) · 1.72 KB
Raw

Project Media Insights

Developed by Darren Colby and Jessup Jong

Get insights into your user inteaction on YouTube. This tool enables you to quickly see trends in comments and sentiments in your videos. We produce time series predictions by day, month, year, and autoregression components using a Bayesian Markov Chain Monte Carlo Prophet, a time series library developed by Facebook. Sentiment classification is based on VADER using preprocessed text. When conducting sentiment classification, the program removes emojis, newlines, tabs, carriage returns, punctuation, and numbers before translating any foreign text to English and correcting misspelled words. This program allows you to visualize the similarity of competitor videos by using the cosine similarity of the GloVe embeddings of their transcripts. Finally, you can use the upload module to automate video uploads.

Features

  • Automate video uploads
  • Can be run by double clicking a bash file
  • Automatic translation of comments and transcripts to English
  • Automatic spelling correction
  • Cosine similarities calculated based on GloVe embeddings
  • Change the forecast period with a slider
  • Uploads secured by OAuth

Data Sources

  • comment data directly from YouTube using GET requests
  • youtube_transcript_api third party library for video transcripts
  • gensim for pretrained GloVe model weights and tokens

Usage

Using the automate.sh script

cd media_insights
automate.sh

Alternatively run by double clicking on automate.sh

Makefile

cd media_insights
make

Demo Version

In addition to running the program locally, you can interact with a live version at https://mediainsights.streamlit.app