lsa-non-incremental-vocabulary-limitation
IN premise — summaries/2026/08/24/wiki-Semantic_similarity.md
Created 2026-08-25T02:58:54+00:00
Latent Semantic Analysis requires a non-incremental vocabulary and has expensive pre-processing, limiting its use in dynamic or real-time settings.
Summary
Latent Semantic Analysis builds its word-to-meaning map from a fixed snapshot of vocabulary and needs a heavy upfront computation pass, so it can't absorb new terms or shift its understanding on the fly. This makes it a poor fit for any system that has to keep learning as the world changes around it, like live conversation, fast-moving news analysis, or evolving product catalogs.