Text Mining Function

1. Differential Relative Rates (DRR)
2. Classifier: Multinominal Naive Bayesian (MNB)
3. Classifier: Weka Naive Bayes
4. Classifier: Decision Tree (DT) Generate Graphic Tree
5. Predict: Multinominal Naive Bayesian (MNB). Train on c1 predict on c2.
6. Vector Space (experiment) Normalize Vectors (LogLinear).
7. Classifier: SVMLight C:
8. Classifier: SMO (Weka)
9. Weka Information Gain

Runname:
Top Feature Display Limit:

or

Options and Configurations

1. Balance instances. (applied to all tests).
   If the two corpora are not balanced by number of documents, performs a random selection from larger set. Important Note: hitting reload on a result from this option will select another test set (randomly), so you will get different results!!!
2. Limit features in less than 100% 95% 90% 85% 80% 75% 70% 65% documents and more than 0% 5% 10% 15% 20% 25% 30% 35% of documents.
3. Stem input words
4. Restrict Document Feature Frequency: LowerBound UpperBound applies to MNB and DT.
   MNB performs better on relatively infrequent words (2-20?). DT splits on ranges, so 2-100 or so may be best. Set to 1 and 5000 to use all but most frequent words.
5. Random Falsification (randomize instances, as a test of classification model).
6. No select or select from bibliography.
   This will allow the user to select/unselect from the two subcorpus selected from the above form (not implemented).

Input Features to use separated by spaces, using full regular expressions:

Input Features to EXCLUDE from other criteria separated by spaces, using full regular expressions: Eg: d j v s c t i p f archit architecture architect

or