Sentimental analysis for YouTube comments
- Find suitable pretrained NLP model that could be applied directly without training (Bert?)
- Youtube comment download feature (APIs)
- Model training and fine-tuning
- Implement download comments from specific Youtube videos
- Preprocessing and data cleaning (nulls check, emoji etc.)
I've finished developing download comments from Youtube. Please see downloader.py for details. Below is the working principle of this youtube comment downloader.py.
python downloader.py --output <output-path> --q <query-word> --maxresults <max-num-of-videos>
Example:
python downloader.py --output res.json --q amazonglacier --maxresults 2
TL;DR: just simply install requirements.txt
pip install requirements.txt
Prerequisites:
- Python Version >= 3.7
- pip
- The Google APIs Client Library for Python:
pip install --upgrade google-api-python-client
- The google-auth, google-auth-oauthlib, and google-auth-httplib2 for user authorization
pip install --upgrade google-auth google-auth-oauthlib google-auth-httplib2
- Requests
pip install requests
- lxml
pip install lxml
- cssselect
pip install cssselect
Resources: Google official Python API client repository:
YouTube Python API:
- Feed the data into the model, compare the output with groundtruth
- Fine-tune parameters (Are we doing binary classification here? i.e. 1 for positive emotion and -1 for negative?)
- Results explanation and investigation
- Write the complete report
- Design slides