You probably want to start with papers by Gerard Salton and work your way forward from there. Basically, a topic model is a vector space model (typically implem

You probably want to start with papers by Gerard Salton and work your way forward from there. Basically, a topic model is a vector space model (typically implemented as a matrix where each row is one vector and each column is a dimension). How the matrix gets filled depends a lot on the underlying data, the assumptions that can be made, and the application.