A practical introduction to information retrieval architecture, written primarily for information literacy and evidence synthesis librarians, with a smaller secondary audience of systems/discovery librarians. It covers Boolean search, BM25, dense retrieval, hybrid search, reranking, query understanding, agentic search, retrieval failure and evaluation.
Read the textbook — version 1.2.1, three parts, fifteen chapters and seven appendices.
The repository also includes two interactive companions: the BM25 Evidence Lab and the Vector Similarity Lab.
search-textbook.html is the sole maintained edition and source of truth. The retired single-flow edition can be recovered from Git history, and its former public URL redirects to the textbook. See tools/README.md for the maintenance passes, and CHANGELOG.md for what changed between versions.
The text, tables, diagrams and code are licensed CC BY 4.0 — copy, excerpt, translate, adapt and redistribute for any purpose, including commercially, with credit. No permission request is needed.
Reproduced research figures (including Figure 5.5 from Mikolov and colleagues) and screenshots of commercial products (scite, Google Scholar, Web of Science and others) are third-party material reproduced for comment and criticism, and are not covered. Product names and trademarks belong to their owners. See LICENSE for the full terms.
Start with the one-hour workshop, Why did this search miss a paper? It includes a prepared case, timed facilitator guidance, worked answers and a two-page participant handout, with no advance reading or subscription access required for participants. The teaching notes also provide audience routes, three longer course shapes, an eight-point assessment rubric, discussion prompts, product-claim verification guidance and a record for improving the next delivery. Workshop timings are provisional pending teaching experience.
Chapters and appendices are written to be assignable on their own, and each has a stable link — use the Copy link control in any chapter heading. Every chapter ends with Check yourself questions whose answers stay hidden until opened. Three combinations stand alone particularly well: Part I (Chapters 1–4) as an information-literacy grounding in retrieval foundations and lexical search; Chapters 2, 11 and 13–15 with Appendix F for evidence synthesis, where Appendix F should be treated as core rather than optional; and Chapter 15 with Appendix D for systems/discovery work on procurement and evaluation.
Tay, A. C. H. (2026). How search decides what you see: A librarian's guide to Boolean search, BM25, embeddings, reranking, and the retrieval pipelines behind hybrid and agentic search (Version 1.2.1). https://aarontaycheehsien.github.io/Information-retrieval-crashcourse/
Generative AI assisted with research, drafting, revision, code and original explanatory illustrations. The author reviewed the outputs, checked cited claims against the linked sources and takes responsibility for the final content. Read the full generative-AI use disclosure.