One of them is: Tishby, N. and Gorin, A., Algebraic learning of
statistical associations for language acquisition, Computer Speech
and Language (1994) 8, 51-78, in which the use of such factors seems
to me to be implicit in the method. Another is: Juergen Schmidhuber
and Daniel Prelinger, Discovering Predictable Classifications,
Neural Computation, 5(4):625-635, 1993, which goes on to use
the factors found to define the class of associated words.
I would appreciate hearing about any other studies, and what other
peoples experience of this approach has been. Any general comments
or opinions would also be welcome.
Rob Freeman
lcrobf@usthk.ust.hk
Language Center,
Hong Kong University of Science and Technology