Teaching & Tools

Open lessons and R packages for research on work, language, and data

Lessons

These lessons are for readers who are new to the topic. Lessons include R examples and output from the site build.

Periodic Table of NLP Tasks in R

This map shows 81 natural language processing (NLP) tasks on one page. Related tasks are in the same column, so you can find a lesson by its position. Each tile opens a lesson. Each lesson starts with a problem and defines new terms where they first appear.

Two parts of the site are not on the map. The guide Using a language model in a reproducible analysis shows how to compare model labels with labels from a person. It also shows how to correct an estimate for label errors. The section Beyond the periodic table uses R and invented test results to compare two versions of a complete system.

81 of 81 lessons published All 6 stages complete
  • Source and training data Encoding, files, APIs, scraping, OCR, annotation 13 / 13
  • Language structure Tokens, morphology, phrases, entities, sentences, documents 29 / 29
  • Models and analysis Training, evaluation, classification, topics, similarity 20 / 20
  • Language generation Next-token prediction, report writing, translation, summarization, paraphrasing, long documents 6 / 6
  • Language systems Relation extraction, question answering, chatbots, semantic search, knowledge bases, e-discovery 6 / 6
  • Visualization Apps, annotated text, word clouds, embeddings, timelines, maps, knowledge graphs 7 / 7
Open the table Start at lesson one Beyond the periodic table Source

R packages

Open-source packages for empirical research and for working with models from R. Each one has reference documentation and a public repository.

huggingfaceR

175 stars on GitHub

Run Hugging Face models from R. Pull pretrained models for classification, embeddings, and generation without leaving a tidyverse workflow.

Documentation Source

onet2r

5 stars on GitHub

A client for O*NET occupational data. Read archived releases, bridge taxonomy changes between vintages, and weight by employment so a cross-section is not mistaken for a time series.

Documentation Source

cmapr

3 stars on GitHub

Tools for analyzing global career mobility data using tidyverse workflows. Build and study the transition networks behind occupational movement.

Documentation Source

foundryR

1 star on GitHub

An R interface to Azure AI Foundry. Call hosted models and manage deployments from an analysis script instead of a separate console.

Documentation Source