Extracted Features Notebooks

Updated May 2023

This repository contains notebooks for retrieving and analyzing extracted features from textual corpora. Its intended use is to analyze a collection of science fiction texts at Temple University which are currently under copyright.

Contents

This repository contains two folders:

• notebooks: Python and R notebooks for retrieving and analyzing extracted features, or non-consumable, disaggregated versions of copyrighted work for research purposes
• data: Sample texts and outputs

Setup

Requires Python 3.9+. Install core dependencies with uv:

uv sync

Optional dependency groups for specific notebooks:

uv sync --extra nlp          # BookNLP + spaCy (feature extraction)
uv sync --extra embeddings   # Word2Vec, Atlas, sentence-transformers
uv sync --extra ocr          # Pytesseract OCR

Or with pip: pip install -e ".[nlp,embeddings,ocr]"

R notebooks require R and RStudio: https://posit.co/download/rstudio-desktop/

Notes

• Files for sectioning/disaggregation must be UTF-8 encoded text (.txt) files
• CSV for LDA topic modeling must contain disaggregated texts; BERTopic works best with aggregated data from which stopwords have NOT been removed
• Parameters (chunk size, number of topics, iterations, passes) are set in the code and documented in comments

Contributors

Jeff Antsen: R notebooks and documentation

Megan Kane: Python notebooks and documentation