The code is organized according to the following directory structure:
clustering_experiments: Jupyter notebook using Python 3.6 for basic data exploration and experiments with k-means clustering on Wikipedia timeseries data
historical_vis: HTML with D3 JS code for visualizing long-term history of a trend’s popularity on Wikipedia and Twitter
real_time_vis: Main Python script, additional helper Python and Node.js scripts and an HTML with D3 JS code for visualizing real-time trends
related_trends: Lib folder with the D3 JS requirements along with an NLP visualization with the main python script to get related trends, and an HTML to visualize them
scripts: 1. Node.js script to get Google Trends interest over time data for the trends in the historical Twitter dataset and 2. Python script to get wikipedia page view data for the trends in the historical Twitter dataset
package.json and package-lock.json: Node.js dependencies
python_requirements.txt: Python package dependencies
First install the NPM package manager and Node.js, then run the following command from the CODE directory to install all Node.js dependencies:
npm install
First install Python 3 and the pip package manager, then run the following command from the Python directory to install all Python dependencies:
pip install -r python_requirements.txt
It’s recommended to install these packages in a virtual environment to avoid any conflicts with your system Python.
Open ts_vis.html in a browser and interact with the visualization by searching for trending topics in the search bar.
To first get the spaCy model, run the following command from the related_trends/NLPVisualization directory:
python -m spacy download en_core_web_md
To get related trends, run the following command from the same directory:
python nlpFunctions.py
From the same directory, open graph.html to visualize the related trends
To access the Twitter API, you need an API key and access token which you can get by signing up for a developer account here: https://developer.twitter.com/en/docs.html. Store your credentials in the following JSON format:
{"consumer_key": "your consumer key","consumer_secret": "your consumer secret key","access_token": "your access token","access_token_secret": "your access secret token"}
To run the visualizations on the current trends, run the following command from the real_time_vis directory:
python real_time_vis.py -i <path-to-twitter-credentials-json>
The script will open an HTML page in your default browser at the end of its execution which will show the visualizations. If your default browser is not Firefox, copy-paste the link to Firefox.
To get the Google Trends interest over time data for the trends in the historical Twitter dataset, run the following command:
node googletrends_interestOT.js <path-to-twitter-csv> <path-to-output-csv>
To get the Wikipedia pageviews, run the following command:
python wiki_collection_multi.py
Expects a csv of twitter trends and associated datetimes to reside in the same directory.
