laito / cleartk

Automatically exported from code.google.com/p/cleartk
0 stars 0 forks source link

consistent naming of DataWriters etc. in cleartk-ml-* modules #319

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
So our naming of binary vs. multi-class classifiers is inconsistent:

BinaryLIBSVMDataWriter.java
MultiClassLIBSVMDataWriter.java

BinaryMalletDataWriter.java
MalletDataWriter.java

BinaryMaxentDataWriter.java
MaxentDataWriter.java

OVASVMlightDataWriter.java
SVMlightDataWriter.java

I think we should be consistent across the different cleartk-ml-* modules. 
Also, for the purposes of auto-complete, it's much nicer if the class starts 
with e.g. LIBSVM or SVMlight. I also think it's misleading to call 
MultiClassLIBSVMDataWriter "MultiClass" since it can just as easily be used 
with a two class problem. Instead, it just happens to be the one that works 
with String outcomes. (Also, you could easily have a multi-class problem where 
the types were integers, but that wouldn't work with MultiClassLIBSVMDataWriter 
unless you converted them to Strings.)

So here's my proposal:

LIBSVMStringOutcomeDataWriter.java
LIBSVMBooleanOutcomeDataWriter.java

MalletStringOutcomeDataWriter
MalletBooleanOutcomeDataWriter

MaxentStringOutcomeDataWriter
MaxentBooleanOutcomeDataWriter

SVMlightStringOutcomeDataWriter
SVMlightBooleanOutcomeDataWriter

We would, of course, want to rename all the ClassifierBuilders, etc. similarly.

We can introduce these names in a backwards compatible way by:

(1) Renaming each class as described above
(2) Creating an empty subclass of the new class that has the old name
(3) Deprecating the subclass with the old name

Original issue reported on code.google.com by steven.b...@gmail.com on 31 Jul 2012 at 9:21

GoogleCodeExporter commented 9 years ago
I like this proposal, but I'm wondering if Maxent* should be called 
OpenNLPMaxent - not that we have any other Maxent libraries (Mallet maybe).

Original comment by lee.becker on 31 Jul 2012 at 9:43

GoogleCodeExporter commented 9 years ago
This issue was closed by revision r3966.

Original comment by steven.b...@gmail.com on 5 Aug 2012 at 7:00