Bag-of-Words Problem and Semantic Analysis in Fock Space
Diederik Aerts, Marek Czachor · arXiv (Cornell University) · 2003
Latent Semantic Analysis (LSA) if reformulated as a Hilbert-space problem reveals formal structures known from quantum mechanics (words and sentences as states, similarity of meaning as transition amplitude, relation between words and sentences analogous to supersymmetry, text as a supercharge). These facts allow to introduce to quantum information theory concepts from quantitative linguistics. Simultaneously, the conceptual and mathematical structures developed within quantum information theory may enrich LSA. In particular, the greatest difficulty of LSA (the "bag-of-words" problem) has a natural solution if one replaces the original semantic space by its Fock space.