NOTE: the instructions contained on this page were written for use specifically in the CLSP WS02 lab.
They may refer to applications, files or other materials that are not accessible from other locations.

DECISION TREES AND DETERMINISTIC CPT SYNTAX
-------------------------------------------

In this document you will find some basics about the master file that you
will need to change during the lab session.


You will need to add at least two structures in the master file.  One
is the decision tree and the the other is the corresponding
DeterminisiticCPT.


1. You will find the following entry in the master file.

    DT_IN_FILE inline
    6 % number of decision trees
 
This indicates that you will find 6 decision trees in the following
text.  For each decision tree there will be an index.  For example:

    5
    endOfDecodedUtterance
    1 % one parent
    0 2 1 default
       -1 1
       -1 0

The first number 5 is the index of the decision tree.
"endOfDecodeUtterence" is the name of the decision tree.  Then you
will see the number of parents, and the tree itself.

    0 2 1 default

It means that you are looking at parent 0, and it has two branches. When 
the value of parent 0 is 1, you go to the first branch otherwise go to
the second branch.

    -1 1
means that it is a leave node (no branches) and the returned value is 1.


2.  When you create a decision tree, you need to map a DeterministicCPT 
to that decision tree.  In the master file you will find section like

    DETERMINISTIC_CPT_IN_FILE inline
    10

which tells you that you will find 10 deterministic conditional
probability tables(CPT's).  Corresponding to the previous decision tree, 
you will find

    7
    endOfDecodedUtterance
    1 % num parents
    2 2
    endOfDecodedUtterance

Here 7 is the index of the DeterministicCPT, and
"endOfDecodeUtterence" is the name of the CPT.  Following that you can
find the number of parents, cardinality of the parents and the
cardinality of itself.  Finally, it tells which decision tree will be
used in the DeterminiticCPT.


In the tutorial, you will add at least one node to each frame and
create DeterministicCPT for them.  You will need to write a decision
tree and a DeterministicCPT that uses this decision tree.