Russian Troll Twitter Activity Classifier

Context

Disinformation and inauthentic coordinated activity require scalable text classification methods that hiring managers can evaluate as applied NLP evidence.

Method / approach

Conducted sentiment analysis and NLP on 200,000+ records using Python (NLTK, scikit-learn) and built a Naive Bayes classifier for troll vs. authentic accounts.

Tools and technologies

• Python
• NLTK
• scikit-learn
• Naive Bayes

Key deliverables

• Classifier and text-as-data analysis workflow

Results / findings

• Reported classification accuracy of 71.9% on the documented evaluation setup.