Browse Prior Art Database

Optimal Phonological Rules for Decision Trees in Continuous Speech Recognition

IP.com Disclosure Number: IPCOM000121779D
Original Publication Date: 1991-Sep-01
Included in the Prior Art Database: 2005-Apr-03
Document File: 2 page(s) / 69K

Publishing Venue

IBM

Related People

Burshtein, D: AUTHOR [+5]

Abstract

We give the polynomial algorithm for finding the optimal rules for decision trees in continuous speech recognition. The optimality criterion for these rules are provided by a special impurity partition function that is derived from the assumption of Poison source for frequencies of acoustic labels in strings.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 72% of the total text.

Optimal Phonological Rules for Decision Trees in Continuous Speech
Recognition

      We give the polynomial algorithm for finding the optimal
rules for decision trees in continuous speech recognition. The
optimality criterion for these rules are provided by a special
impurity partition function that is derived from the assumption of
Poison source for frequencies of acoustic labels in strings.

      In [1] the special decision tree with phonological rules in
continuous speech was introduced.  In this tree questions are chosen
to maximize the purity function.
where the products are taken over N = NL + NR strings  sj . (ML
and MR denote models at the nodes of the tree that corresponds to the
split of the set of strings into subsets L and R).  Assuming the
Poison source for frequencies of acoustic labels in strings [1], the
problem (1) becomes equivalent to the problem of finding partition of
the all considered set of strings into subsets L,R of strings that
maximize the impurity function
where F is the number of different labels (that can occur in
strings), sj = count of label i in the string s.  The exhaustive
search through all possibilities for finding the optimal split is
exponential in N.  In this invention we describe a method for solving
this problem that is polynomial in N.  The method is based on
establishing the equivalence of the problem (2) to the problem of
minimizing the impurity partitions of a special type for which such
algorithms are known from...