Classifying Fanfiction from Archive of Our Own (Ao3) as either First, Second, or Third Person using Statistical/Machine Learning Algorithms

• Notebook on Data Extraction
  • Used Archivist.py to scrape Work Id's and first chapter tests from Ao3
• Notebook on Cleaning & Feature Engineering
  • Removed html artifacts from text.
  • Tokenized by words and by sentences.
  • Created features based on frequency ratios for key words that 1. appear anywhere 2. begin a sentence
• Notebook on training ML models
  • Logistic Regression: 86% Accuracy
  • K Nearest Neighbors: 93% Accuracy
  • 

• Implement more models:
  • Random Forest
  • Boosting
• Productionize model for a browser extention for Ao3 users
• Investigate Features that might allow the model to differentiate between Third Person Subtypes (i.e. Limited, Omniscient)