ADR-011: information-retrieval joins the topics

Context

On 2026-09-27 the owner asked the record to file C. J. van Rijsbergen, The Geometry of Information Retrieval (Cambridge, 2004).

The book recasts information retrieval’s vector-space, probabilistic and logical models in the language of Hilbert space. Documents are vectors. Relevance and aboutness are observables. Retrieval is measurement, with a non-Boolean logic of projectors.

The words the vocabulary had describe its formalism, not its subject:

• mathematics;
• quantum-foundations;
• logic;
• probabilistic-modeling.

The seed named the gap (wants: information-retrieval). The record’s rule is to add the word rather than tag with the nearest wrong one (ADR-006, ADR-007, ADR-009, ADR-010).

One held work already needs the word as well. Gebhart, Hansen & Schrater’s Knowledge Sheaves embeds knowledge graphs in order to answer multi-hop queries.

Decision

Add information-retrieval to topics, with a label and a blurb. It does not join the philosophy group.

work	tags
van Rijsbergen, The Geometry of Information Retrieval	information-retrieval first, then mathematics, quantum-foundations, logic
Gebhart, Hansen & Schrater, Knowledge Sheaves	information-retrieval added

The book’s seed also proposed probabilistic-modeling. It is not carried. The book’s probability is quantum probability over projectors, which quantum-foundations covers. The topic’s blurb, Bayesian inference and statistical models, would mislead a reader browsing it.

Alternatives

• Tag the book by its formalism only. A reader browsing for retrieval would never find it. A reader browsing quantum-foundations would find a book about search engines.
• Fold retrieval into representation-learning. Retrieval predates learned representations. Most of the book’s models are not learned at all.