> Some people pointed me to "A Tutorial on Hidden Markov Models and
> Selected Application in Speech Recognition" by Lawrence R. Rabiner,
> which originally appeared in the Proceedings of the IEEE, Vol. 77,
> No. 2, pages 257-286. It is also available as a reprint on pages
> 267-296 in "Readings in Speech Recognition", ed. by Alex Waibel and
> Kai-Fu Lee, San Mateo: Morgan Kaufman, 1990. This tutorial is in fact
> an excellent introduction into HMMs and their applications for speech
> recognition.
We cannot fully agree with this high estimation of the Rabiner's paper. On the o
ne hand
it is indeed an excellent introduction or tutorial, for it is written in a style
that is
easily comprehensible for the mathematical layman with some formal training.
On the other hand, there are some errata in the formulae presented in the paper.
And those you
can only be found by someone with a real mathematical background. Errata of this
kind are
the more dangerous if you encounter them in a TUTORIAL which is addressed to the
layman with some formal training (as I myself am).
We have written on occasion of reviewing a tagger which is based on HMM-techniq
ues a
paper which adds the necessary corrections to Rabiner's formulae.
With the Rabiner paper in hand, you can use this paper as a "patch" to his formu
lae set.
My colleague Holger Trebbe wrote the paper in the laconic style which seems to b
e
usual in the math community.
You find the paper in the WWW:
http://santana.uni-muenster.de/Linguistik/papers/eerrata.ps
In the papers directory, you also find a general description of our approach to
tagging
Lothar Lemnitzer