Aspect model

From Cohen Courses
Jump to navigationJump to search

Aspect model is a latent variable model for general co-occurrence data which associates an unobserved class variable z with each observation that is in the form of co-occurrence between a word w with a document d, i.e. an observed pair (w, d).

The model is based on two assumptions:

  • Observation pairs (w, d) are assumed to be generated independently
  • Conditioned on the latent class z, words w are generated independently of the document d.

Discussed in Class Meeting for 10-802 02/22/2011

Taken from the paper: Thomas Hofmann, Probabilistic Latent Semantic Indexing, SIGIR 2009

Relevant Papers