Browse Prior Art Database

Confusion Matrix Generator

IP.com Disclosure Number: IPCOM000106858D
Original Publication Date: 1993-Dec-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 4 page(s) / 167K

Publishing Venue

IBM

Related People

Kennedy, PJ: AUTHOR [+3]

Abstract

Disclosed is a program for providing, in matrix form, the distribution of character recognitions and miss-recognitions provided by an electronic system for writing recognition. A confusion matrix shows the number of times a character has been fed into the recognition system, together with the numerical distribution of characters it was recognized to be.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 33% of the total text.

Confusion Matrix Generator

      Disclosed is a program for providing, in matrix form, the
distribution of character recognitions and miss-recognitions provided
by an electronic system for writing recognition.  A confusion matrix
shows the number of times a character has been fed into the
recognition system, together with the numerical distribution of
characters it was recognized to be.

      This program is a result of the Recognition Accuracy Project
(RAP), used to evaluate the accuracy of handwriting recognition
algorithms developed to run in a pen-point, or tablet computer
environment.  Data on various strokes used to write characters,
collected by the RAP software, is fed back through recognition
algorithms to determine how incremental changes to those algorithms
effect the overall accuracy of the recognition system.  Information
collected in this way is displayed in the form of a confusion matrix.
In this way, a confusion matrix is generated using either a RAP file
based on handwriting samples from a single subject, or a summation
file created from two or more RAP files to evaluate handwriting
samples from two or more subjects.

      The output of this program is an ASCII file containing the
results in a matrix format.  The record length of these files varies
with the number of instances of characters being tested, as each
column in the matrix is able to contain the largest number of
instances of any character in the set being tested.  Another form of
output is a summation file, which is made from the records of
multiple RAP subject files.

      In the initialization phase, the first dimension of a
two-dimensional array is set with the ASCII codes of the character
set that has been fed into the recognition system.  The order in
which codes are put into the array determines the order that their
corresponding characters will appear in the matrix.

      In the first display screen, the program is identified by a
boxed comment in the center of the screen, and three function
keys--F2, F4, and ESC--are defined in a menu bar extending across the
top of the screen.  The F2 key is defined as meaning "Create a Matrix
File."  When this key is depressed, a prompt for an input file name
is displayed.  This file name must identify either a RAP generated
subject file or a summation file.  When a valid input file name has
been entered, the system displays a prompt for an output file name,
which can be any valid DOS file name.  When the output file name is
entered, the system reads the input file, the data is stored in the
form of an array, the array is scanned, and output records are
created and written to the output file.  The F4 key is defined as
meaning "Output Options."  When this key is depressed, the second
display screen is presented.  In the first display screen, the ESC
key is defined as meaning "Exit Program."  When this key is
depressed, the system returns to the DOS prompt.

      In the second, "O...