The Infinite Hierarchical Hidden Markov Model
Katherine Heller, Yee Whye Teh and Dilan Gorur
In: AISTATS 2009, Florida, USA(2009).
In this paper we present the Infinite Hier- archical Hidden Markov Model (IHHMM), a nonparametric generalization of Hierarchical Hidden Markov Models (HHMMs). HHMMs have been used for modeling sequential data in applications such as speech recognition, detecting topic transitions in video and ex- tracting information from text. The IHHMM provides more flexible modeling of sequen- tial data by allowing a potentially unbounded number of levels in the hierarchy, instead of requiring the specification of a fixed hierar- chy depth. Inference and learning are per- formed efficiently using Gibbs sampling and a modified forward-backtrack algorithm. We present encouraging results on toy sequences and English text data.