| head_bayes.js |
|
876 |
| resources |
|
|
| test_bug228675.js |
import-globals-from resources/trainingfile.js |
3948 |
| test_customTokenization.js |
|
5041 |
| test_junkAsTraits.js |
|
15972 |
| test_msgCorpus.js |
msgCorpus.corpusCounts(1001, messageCountObject);
let v1001 = messageCountObject.value;
msgCorpus.corpusCounts(1003, messageCountObject);
let v1003 = messageCountObject.value;
dump("pre-clear value " + v1001 + " " + v1003 + "\n");
/* |
4675 |
| test_traitAliases.js |
These tests rely on data stored in a file, with the same format as traits.dat,
that was trained in the following manner. There are two training messages,
included here as files aliases1.eml and aliases2.eml Aliases.dat was trained on
each of these messages, for different trait indices, as follows, with
columns showing the training count for each trait index:
file count(1001) count(1005) count(1007) count(1009)
aliases1.eml 1 0 2 0
aliases2.eml 0 1 0 1
There is also a third email file, aliases3.eml, which combines tokens
from aliases1.eml and aliases2.eml
The goal here is to demonstrate that traits 1001 and 1007, and traits
1005 and 1009, can be combined using aliases. We classify messages with
trait 1001 as the PRO trait, and 1005 as the ANTI trait.
With these characteristics, I've run a trait analysis without aliases, and
determined that the following is the correct percentage results from the
analysis for each message. "Train11" means that the training was 1 pro count
from aliases1.eml, and 1 anti count from alias2.eml. "Train32" is 3 pro counts,
and 2 anti counts.
percentage
file Train11 Train32
alias1.eml 92 98
alias2.eml 8 3
alias3.eml 50 53
|
4947 |
| test_traits.js |
|
8229 |
| xpcshell.ini |
|
199 |