-
So far only the classifiers that extends SemiSupervisedClassifierLearner or SemiSupervisedBatchClassifierLearner can be used in a semi-supervised way - unfortunately minorthird does not have a lot of these.
2009-12-09 00:09:07 UTC in MinorThird
-
Yeah, some of these build targets are out-dated, but we've left them in in case there are still people out there using these. We may remove them in the future.
Minorthird has definitely been used for topic classification in general, but I am not sure about sentiment analysis specifically. You can surely use Minorthird out-of-the-box if you pose sentiment analysis as a sentence-level document...
2009-12-04 04:52:18 UTC in MinorThird
-
The build-apps and and build-names targets were specific to some research done years ago; most minorthird functionalities, documented or otherwise, can be built simply with "ant build". Are you using some specific functions that require the classes under ./apps?.
2009-12-04 00:19:21 UTC in MinorThird
-
1 - By default, defDict should be case-insensitive unless you specify "+case". So if you do
defDict +case myDict = one, two;
The dictionary myDict would match "one" and "two" but not "One" or "Two". If you do
defDict myDict = one, two;
Then myDict should match all casings of "one" and "two".
2 - Can you post a sample of your mixup that causes the problem?.
2009-11-25 19:29:13 UTC in MinorThird
-
linfrank changed the public information on the MinorThird project.
2009-10-01 18:05:48 UTC in MinorThird
-
linfrank changed the public information on the MinorThird project.
2009-10-01 18:00:28 UTC in MinorThird
-
All MinorThird documentation for now on will live on the Trac wiki hosted by SourceForge, to facilitate update and addition to the documentation and tutorials. The wiki is at:
https://sourceforge.net/apps/trac/minorthird/wiki
And minorthird.sourceforge.net redirects there.
Old documentation will still be available...
2009-10-01 17:56:47 UTC in MinorThird
-
We've moved the source code a CVS to a subversion repository! The old CVS repository will still be available for a while, but will no longer be updated. Check out the newest version using an SVN client from:
https://minorthird.svn.sourceforge.net/svnroot/minorthird/trunk
Don't forget /trunk -- otherwise...
2009-10-01 17:52:01 UTC in MinorThird
-
Hi Natlu,
How many training examples are needed to produce a good result (say, 80% precision and recall) usually depends on 1) how "difficult" it is to recognize this entity; i.e., the number of features/feature weights needed - this you noted when you realize you need a variety of data that would cover different contexts and forms of this entity; and 2) your feature extractor...
2009-10-01 16:36:38 UTC in MinorThird
-
Hi Haim,
Sorry for the lack of reply - I was away for the summer. Please let me know if you still need help on this.
-frank.
2009-10-01 04:55:28 UTC in MinorThird