Where to Make Your Changes
To ensure that your code works properly not only when run from the command line but also when executed via the submit process, you should limit your changes to addExample() and classify() since these are the hooks that Submit uses to exercise your code. You're free to add other elements further invoked from addExample() or classify(), but note well that main() is NOT executed by Submit so you cannot rely on anything added there.

Changes beyond addExample() and classify(), if you choose to make any, should be done with caution. For java specifically, Submit.java makes calls to getTrainSplits(), train(), readTest(), and evaluate(). Submit also relies on the definition of TrainSplit and directly manipulates boolean FILTER_STOP_WORDS. Note that classify() is called both directly by Submit.java and also from within evaluate(). (As well as from main(), of course.)

For python, Submit.py calls the following: crossValidationSplits(), trainSplit(), train(), test(). Submit also relies on the definitions of classes TrainSplit and Example and directly manipulates boolean FILTER_STOP_WORDS.

Evaluation
Your classifier will be evaluated on two different data sets—both imdb1 mentioned above and a second held-out test set—and each of these is then in turn evaluated twice, once with and once without invoking stop-word filtering.

Running the code
For Java, execute

$ cd java
$ mkdir classes
$ javac -d classes -target 1.6 *.java
$ java -cp classes NaiveBayes ../data/imdb1
This will train the language models for each cross-validation and output their performance.

Note that the directory argument allows you to train on other data sets in the process of your development, but for submission, the Submit.java script expects to be able to find the imdb1 directory in its default location noted here.

Adding a -f flag...

$ java -cp classes NaiveBayes -f ../data/imdb1
...invokes the stop-word filering.

Similarly, for python, execute

$ cd python
$ python NaiveBayes.py (-f) ../data/imdb1
If you're curious how your classifier peforms on separate training and test data sets, you can specify a second directory, in which case the program should train on the entirety of the first set (i.e., without cross-validation) then classify the entire held-out second set.

$ java -cp classes NaiveBayes (-f) train test
or

$ python NaiveBayes.py (-f) train test
Submitting Your Solution
For each language, we have provided a submission script/program which will run your program and classify both imdb1 (via 10-fold cross-validation) and a second unseen data set, both with and without stop-word filtering. Your classifier's decisions are sent to our servers where they are stored and graded once the assignment is due. For each language, the submission program will programmatically call your script and then prompt you for your username and a password. The password can be found at the top of the Assignments page on this website. It will also ask you about which parts of the assignment you wish to submit. For this assignment your system will be evaluated in four parts: each of the two data sets both with and without stop-word filtering.

Here are the basic shell commands which you will need to execute for Java

$ cd java
$ mkdir classes
$ javac -d classes *.java
$ java -cp classes Submit
and for Python

$cd python
$python submit.py
It is important that these scripts be executed from either the java or python directories so that they can easily import the code you've written, but also—and this is critical—so that the relative path "../data/" correctly points to the training and development data (it is hard-coded into the submission scripts).