DarkGeekMS/stylometry-classification

A machine learning project for classifying authors based on their writings.

Stylometry Using Machine Learning

A machine learning project for classifying authors based on their writings.

Dataset

The dataset contains the writings of three authors :

• Edgar Allan Poe (EAP) : 7900 phrases.
• HP Lovecraft (HPL) : 5635 phrases.
• Mary Wollstonecraft Shelley (MWS) : 6044 phrases.

Work Pipeline

• Text Processing :

  • Tokenization (converting into words or tokens).
  • Stemming / Lemmatization (normalizing tokens).

• Feature Extraction :

  • Classical : bag of words / n-grams / TF-IDF.
  • Deep Learning : word embeddings.

• Classification (better approach to train a binary classifier for each author to improve generalization) :

  • Classical : Linear Regression / Naive Bayes (BEST) / SVM / XGBoost.
  • Deep Learning : RNNs.

Installation

• Install requirements.txt using PyPi:

  pip3 install -r requirements.txt

Usage

• For training linear classifier model :

  • Create models folder.

  • Edit configs/lc_config.json.

  • Run train.py :

    python train.py train-lc

• For training neural network model :

  • Create models and w2v_models folders.

  • Download GloVe embeddings and extract it into w2v_models folder.

  • Edit configs/nn_config.json.

  • Run train.py :

    python train.py train-nn

• For inference on linear classifier model :

  python evaluate.py eval-lc --author1 /path/to/author1/text --author2 /path/to/author2/text --model /path/to/model/file

• For inference on neural network model :

  python evaluate.py eval-nn --author1 /path/to/author1/text --author2 /path/to/author2/text --model /path/to/model/file --w2v_path /path/to/w2v/model