This repository contains all the code relating to my research project supported by Warwick Manufacturing Group (WMG).
Note that the video in the downloads folder is simply a placeholder - please contact me (samaksh.agarwal@warwick.ac.uk) if you would like the complete dataset.
The research title was: Factors Affecting the Popularity of VR-Related Educational Videos on TikTok
The data collection stage involved identifying relevant videos using general keyword search terms (e.g. virtual reality learning), and then manually downloading this set of videos from the TikTok website.
The speech and video content of these videos was then analysed through a data pipeline implemented in Python.
The speech/text analysis pipeline for each video included the following stages:
- Video Transcription (using the pydub and speech_recognition libraries)
- Punctuation Restoration (using the rpunct library)
- Keyword Extraction (using the yake library)
FFmpeg (required for video transcription) can be downloaded here: https://ffmpeg.org/download.html
Furthermore, the transitions in each video were analysed in order to record the number of distinct scenes and calculate the average scene length (this was done using the pyscenedetect library).
Further statistical analysis and data visualisation was performed using dedicated libraries such as pandas, NumPy, and matplotlib.
Using the results of this data analysis, an academic research paper was written. The paper was accepted by the AIKIEE-2023 conference and has been submitted to the IEEE Xplore Digital Library for potential publication.
The final paper can be found at: https://drive.google.com/file/d/12nb_i909t0K-BJZQlK7t7mSTReOGsj9s/