Random indexing

An incremental dimensionality-reduction method that assigns sparse random index vectors to items and accumulates their contextual vectors, approximating high-dimensional distributional geometry in fixed space.

Version

v1 · 2026-09-08 · History

Domain-specific #

6389

Origin domain

distributional semantics

Subdomain

random projection embeddings

Core Idea

Random indexing represents each context by a sparse near-orthogonal random vector and represents an item by summing the vectors of contexts in which it occurs. Johnson-Lindenstrauss-style random projection approximately preserves pairwise geometry while incremental addition avoids constructing or factorizing a full term-context matrix. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

The load-bearing residual is not the broad topic of distributional semantics. It is online random-projection construction of distributional representations without a growing explicit context dimension.

Scope of Application

Random indexing belongs to distributional semantics and is useful where the analyst can specify items and contexts, high-dimensional sparse random index vectors, a fixed lower-dimensional space, weighted co-occurrence events, accumulated context vectors, and a similarity measure, then evaluate index vectors are generated independently under a declared sparse distribution and semantic vectors accumulate co-occurrence contributions in one fixed dimensional space. The scope is broad within that domain but bounded by the need for index vectors are generated independently under a declared sparse distribution and semantic vectors accumulate co-occurrence contributions in one fixed dimensional space. The entry records a descriptive analytical identity; practical use requires the governing domain's evidence, standards, and safety obligations.

Clarity

The abstraction clarifies a crowded vocabulary by making index vectors are generated independently under a declared sparse distribution and semantic vectors accumulate co-occurrence contributions in one fixed dimensional space the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test. A bare label is insufficient because the name Random indexing can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated.

Manages Complexity

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Random indexing. Random indexing compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning

1. Identify the carrier. State what the elements, states, objects, or observations are: items and contexts, high-dimensional sparse random index vectors, a fixed lower-dimensional space, weighted co-occurrence events, accumulated context vectors, and a similarity measure. Reject examples whose alleged carrier belongs to a different problem. 2. Lock the constitutive rule. Express index vectors are generated independently under a declared sparse distribution and semantic vectors accumulate co-occurrence contributions in one fixed dimensional space independently of one notation or implementation.

Knowledge Transfer

Knowledge transfers strongly among subfields of distributional semantics because they reuse items and contexts, high-dimensional sparse random index vectors, a fixed lower-dimensional space, weighted co-occurrence events, accumulated context vectors, and a similarity measure, Johnson-Lindenstrauss-style random projection approximately preserves pairwise geometry while incremental addition avoids constructing or factorizing a full term-context matrix., and type the carrier, state every parameter and convention in the definition, test that index vectors are generated independently under a declared sparse distribution and semantic vectors accumulate co-occurrence contributions in one fixed dimensional space, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions

Current abstraction Random indexing Domain-specific

Parents (1) — more general patterns this builds on

• Random indexing is a kind of Dimensionality Reduction Prime

  The proposed strict upward parent is prime:dimensionality_reduction.

Hierarchy paths (4) — routes to 3 parentless roots

• Random indexing → Dimensionality Reduction → Approximation → Representation → Abstraction

Show alternative paths (3)

Neighborhood in Abstraction Space

Random indexing sits in a sparse region of the domain-specific corpus (62^nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Multivariate & Spatial Statistics (13 abstractions)

Nearest neighbors

• Word embedding — 0.86
• Euclidean random matrix — 0.86
• Krichevsky–Trofimov estimator — 0.86
• Multivariate t-distribution — 0.86
• Maximal information coefficient — 0.85

Computed from structural-signature embeddings · 2026-09-08