Natural Language Processing

These tutorials provide a hands-on introduction to analyzing text data for research, covering the full spectrum from deterministic rule-based methods to modern LLMs.

The Research Reproducibility Triad

Before diving in, keep these principles in mind for all text analysis work:

1. Automate — transition from interactive notebooks to scripts designed for batch processing

2. Test — embed sanity checks and unit tests at every step to validate extraction accuracy

3. Abstract — separate core logic (your functions) from specific parameters (model versions, file paths)

Tutorial Overview

Tutorial	Methods Covered	When to Use
1. Deterministic Extraction	Regex, tokenization, stemming, POS tagging	Pattern-based extraction; maximum reproducibility
2. Quantifying Features	Dictionary methods, TF-IDF	Measuring known constructs; regression inputs
3. Paid LLMs	OpenAI, Anthropic, Google APIs	Nuanced classification; when context matters
4. Open-Source LLMs	Ollama on KLC	Sensitive data; cost at scale; reproducibility
5. Validation and Rigor	Cohen’s Kappa, cross-validation, Gold Standard	Validating results for academic publication

For interpretive coding with rubrics and adjudication, see Qualitative Coding with LLMs (separate walkthrough).