Tk. Landauer et St. Dumais, A SOLUTION TO PLATES PROBLEM - THE LATENT SEMANTIC ANALYSIS THEORY OFACQUISITION, INDUCTION, AND REPRESENTATION OF KNOWLEDGE, Psychological review, 104(2), 1997, pp. 211-240
How do people know as much as they do with as little information as th
ey get? The problem takes many forms; learning vocabulary from text is
an especially dramatic and convenient case for research. A new genera
l theory of acquired similarity and knowledge representation, latent s
emantic analysis (LSA), is presented and used to successfully simulate
such learning and several other psycholinguistic phenomena. By induci
ng global knowledge indirectly from local co-occurrence data in a larg
e body of representative text, LSA acquired knowledge about the full v
ocabulary of English at a comparable rate to schoolchildren. LSA uses
no prior linguistic or perceptual similarity knowledge; it is based so
lely on a general mathematical learning method that achieves powerful
inductive effects by extracting the right number of dimensions (e.g.,
300) to represent objects and contexts. Relations to other theories, p
henomena, and problems are sketched.