lsa-svd-term-document-matrix

IN premise — summaries/2026/08/24/wiki-Semantic_similarity-chunk-2.md

Created 2026-08-25T02:58:53+00:00

Latent Semantic Analysis (LSA; Landauer & Dumais, 1997) computes similarity via Singular Value Decomposition (SVD) of a term–document matrix, addressing synonymy and polysemy simultaneously.

Summary

Latent Semantic Analysis finds hidden topic structures inside a word-by-document table by breaking it down with SVD, which lets the system treat related words as interchangeable and disambiguate words that carry multiple meanings. This is a foundational assumption the system leans on whenever it needs to judge whether two pieces of text are semantically related.