As of March 21st 2022, we've tracked at least 602 companies from 50 countries who've announced sanctions in response to Russia's invasion of Ukraine. The complete dataset is available for download free at diffbot.com.

We've since paused real-time tracking as daily new sanctions have fallen tremendously, but you can rebuild it for yourself or extend the app to track any kind of market monitoring signal you like.

Feel free to start here or follow the step by step tutorial.

Requirements

• Python 3.8+
• Diffbot API token (get a trial token here)

Get Started

1. Clone the repository with git clone

2. Navigate into the root directory with cd Sanctions

3. Install fasttext and pandas with

   python3 -m pip install fasttext pandas

4. Create a settings.json file with the following data

   {
   	"token": "YOUR DIFFBOT TOKEN"
   }

5. Follow the steps below to run each script

Step 1 — Get Articles

Download all the articles from the Diffbot Knowledge Graph that match a DQL query targeting "high likelihood of sanctions" articles into a file called articles.jsonl. Depending on the date range, this can be several thousands of articles and take some time.

💡 Skip this step if you're on a trial token or you will blow through your trial limits. A copy of articles.jsonl covering the first 7 days after the day of Russia's invasion is provided in sanctions/.

python3 1_get_articles.py sanctions/

Step 2 — Identify Organizations

Process every article in sanctions/articles.jsonl through NLP API, generating a list of sentences mentioning organizations that we can later classify. The resulting data is saved as facts.jsonl.

💡 Skip this step if on a trial token. A copy of facts.jsonl is also included in sanctions/.

python3 2_call_nl.py sanctions/

Step 3 — Prepare Training

Samples training and validation datasets from cross-referencing facts.jsonl with known sanctions in ground_truth.tsv for training and evaluating our machine learning classifier.

python3 3_prepare_training.py sanctions/

Step 4 — Train Model

Trains the classifier model using the training and validation datasets. Outputs a precision and recall we can use to gauge effectiveness of the classifier.

python3 4_train_model.py sanctions/

Step 5 - Generate Data

Runs the classifier model across the rest of the articles to generate a dataset of sanctions. Also enhances each organization with firmographics from the Diffbot Knowledge Graph for better context.

python3 5_generate.py sanctions/