Example (message filtering system for e-mail)
Problem: file under 'D. Fleming' or 'Hockey' folder?
Feature |
From |
Cc |
Date |
Subject |
Value |
D. Fleming |
None |
October 26 |
Hockey |
Weighting of features that determine action:
Feature |
From |
Cc |
Date |
Subject |
Weighting |
0.90 |
0.08 |
0.01 |
0.88 |
Problem: both weightings very close together (<0.1);
low confidence value
Library of close matches:
(Distance of feature to weighting)
Feature |
From |
Cc |
Date |
Subject |
Value |
D.Fleming |
None |
Oct 11 |
Habs |
Distance |
0 |
0 |
0.9 |
0.98 |
Distance overall |
0.8714
|
Action |
File under Fleming |
Feature |
From |
Cc |
Date |
Subject |
Value |
D. Fleming |
None |
Oct 3 |
Hi |
Distance |
0 |
0 |
0.92 |
1 |
Distance overall |
0.8892
|
Action |
File under Fleming |
Feature |
From |
Cc |
Date |
Subject |
Value |
O. Barnhill |
None |
Oct 7 |
Hockey |
Distance |
1 |
0 |
0.86 |
0 |
Distance overall |
0.9086
|
Action |
File under Hockey |
Feature |
From |
Cc |
Date |
Subject |
Value |
S. Fillmore |
None |
Oct 23 |
Hockey |
Distance |
1 |
0 |
0.9 |
0 |
Distance overall |
0.9090
|
Action |
File under Hockey |