Aspect Based Sentiment Analysis (ABSA) of YouTube Comments!
This follows from my work on my undergrad project for Aspect Based Sentiment Analysis. YouTube Analytics is great but it does not provide a facility for tracking viewer sentiments about a video, specifically which aspects of the video do viewers have strong sentiments about. In this project, I run ABSA on the comments thread for Sirav Raval's awesome YouTube video.
Data gathered using the YouTube Data API's commentThreads.list function.
In one line: Data -> Cleaning -> Feature Extraction -> Pruning -> Frequent Feature Selection -> Sentence Extraction -> Sentiment Analysis!
Download these dependencies using pip:
- TextBlob
- matplotlib
- seaborn
- nltk (download the nltk Stop Words corpus by running nltk.download() in a python shell.)
Run the notebook by typing jupyter notebook in the terminal to run the code in a popup browser. If you don't already have jupyter, download it here.
- Run TF-IDF to extract important terms from the data.
- Use scikit-learn Feature Extraction tools.
- Replace TextBlob Sentiment Scoring with a Deep Learning model.