unnathi10 / hatespeechdetection Goto Github PK

View Code? Open in Web Editor NEW

Detecting hate speech using the spoken content of videos using Machine Learning

Jupyter Notebook 99.64% Python 0.36%

hatespeechdetection's Introduction

Detection of Hate Speech Detection in Videos Using Machine Learning

[Master's Project]

Objective:

To detect hate speech in videos based on the spoken content of the videos using machine learning.

Approach:

Extract audio from video using FFmpeg API
Convert audio to text transcript using Google Cloud Speech-to-Text API
Extract features to get word counts, frequency of word counts, uni-grams and bi-grams
Train Naive Bayes, Linear SVM, Random Forrest and RNN models
Compute Accuracy, Precision score, Recall score, F1 score

Dataset:

YouTube videos are searched and downloaded using a YouTube crawler
Videos are searched using the 'search by keyword' function provided by YouTube Data API
Searched videos are downloaded using Pytube library
Each video is labeled as Normal or Hateful (Racist, Sexist)

Language Used:

Python

Tools Used:

YouTube Data API
FFmpeg API
Google Cloud Speech-to-Text API
Google Storage Bucket
Jupyter Notebook

Note: You can view .ipynb files using nbviewer - https://nbviewer.jupyter.org/

hatespeechdetection's People

Contributors

Stargazers

hatespeechdetection's Issues

Access to raw video data

Hi Unnathi, Good to see your work in the area of hate speech detection from videos. This is emerging area now considering video content on social website. I am also exploring similar topics for my research work. It would be really helpful, if you can share raw video data collected by you so that we can explore other options like audio analysis along with text

Recommend Projects