System and method for text tagging and segmentation using a generative/discriminative hybrid hidden markov model

Abstract

A method for sequence tagging medical patient records includes providing a labeled corpus of sentences taken from a set of medical records, initializing generative parameters θ and discriminative parameters {tilde over (θ)}, providing a functional LL−C×Penalty, where LL is a log-likelihood function where em=1−Σp(x|y), e{tilde over (m)}=1−Σ{tilde over (p)}(x|y) are emission probability constraints, tr=1−Σp(y|y), t{tilde over (r)}=1−Σ{tilde over (p)}(y|y) are transition probability constraints, and extracting gradients of LL−C×Penalty with respect to the transition and emission probabilities and solving θ*,{tilde over (θ)}*that maximize LL−C×Penalty, initializing a new iteration with θ*,{tilde over (θ)}* and incrementing C and repeating until solutions have converged, where parameters θ,{tilde over (θ)} are the probabilities that a new sentence X' is labeled as Y′.