Session 1: Introduction to LLMs as Tools for Linguistic Data Analysis

Day 1 · Introduction & First Experience (1-1)

Learning objectives

By the end of this session you will be able to:

  • Describe, at a high level, what a Large Language Model is and how it is trained.
  • Explain what an embedding is, and how a model comes to represent a word’s meaning in context.
  • Say what an LLM has already been used for in linguistics, and where it has been shown to struggle.

Agenda

  1. Why this course — LLMs are now common tools, and they are already changing how science is written. Annotation is the promising and expensive application, so the question is how to judge whether an LLM is doing it well.
  2. Self-introductions — your name, your research interest, and why you took this course.
  3. Syllabus — the course objectives, the 5-day map, and how you are assessed.
  4. What an LLM is — what you already know about them, a 7-minute overview video, and the three questions we come back to: what the model predicts, where its numbers come from, and what attention does.
  5. A brief history of embeddings — vectors and the RGB colour space, the distributional hypothesis, co-occurrence counts, Latent Semantic Analysis, word2vec, and finally the Transformer and self-attention. This is how text becomes numbers, and how one word gets a different representation in each context.
  6. What LLM can be used, and where it fails — what an LLM lets you annotate that earlier methods could not, why sounding right is not being right, what people have already published using LLMs as annotators, and which part of NLP this course covers (evaluation, not training).
  7. Discussion — what was new, and whether you come out of it optimistic or skeptical.

Bring a laptop today. This session is lecture and discussion, but Sessions 2 and 3 this afternoon are hands-on in Google Colab. If you have not set Colab up yet, work through the Google Colab setup guide before Session 2.

Watch

  • Sanderson, G. (3Blue1Brown). Large Language Models explained briefly (7 min). Lesson page · YouTube
    • We play this in class. Watching it beforehand is optional and makes the session easier to follow. It covers what the model predicts, where its numbers come from, and what attention does.

Reading

Skim (see all course readings):

  • Abdurahman, S., Ziabari, A. S., Moore, A. K., Bartels, D. M., & Dehghani, M. (2025). A primer for evaluating large language models in social-science research. Advances in Methods and Practices in Psychological Science, 8(2). https://doi.org/10.1177/25152459251325174

Slides & Colab