The information of the input source/target word corpora:
======================================================================================
Directory + name of the source corpus  (e.g. ./data/source.fr).
SourceCorpusFile = 
Directory + name of the target corpus  (e.g. ./data/target.en).
TargetCorpusFile = 
======================================================================================


The information of the input source/target word class dictionaries:
======================================================================================
Directory + name of the source word class dictionary (e.g. ./corpus/fr.vcb.classes).
SourceWordClassFile = 
Directory + name of the target word class dictionary (e.g. ./corpus/en.vcb.classes).
TargetWordClassFile = 
======================================================================================


For extracting the phrase pairs with their reordering distances, please define:
======================================================================================
Directory + name of the word alignment file (e.g. ./model/aligned.grow-diag-final-and).
alignmentFile = 
Directory + name of the (output) phrase pairs table (e.g. ./data/extractPhraseTrainTable).
phraseTableFile = 
(Optional) Directory + name of the test corpus (if filled, only extract the phrases from the training corpora that appear in this file).
TestFileName = 
======================================================================================


For generating the phrase reordering probabilities with structured learning, please define:
======================================================================================
Directory + name of the (input) phrase translation table (recommended using filtered Moses's phrase table) (e.g. ./model/phrase-table).
phraseTranslationTable = 
Extract the top N translations for a source phrase (based on the probabilities by Moses's phrase table). 
maxTranslations = 100 
If the phrase translation table is already filtered, fill 1 and 0 otherwise.
tableFilterLabel = 1 
Directory + name of the (output) weight parameter matrix (e.g. ./data/weightMatrix).
weightMatrixFile = 
Need to train the weight matrix (If you have trained the weight matrix, then use 0)? 1. Yes, 0. No. 
weightMatrixTrainLabel = 1
Directory + name of the (output) sentence phrase options table (e.g. ./data/phraseOption).
phraseOptionFile = 
Directory + name of the test corpus (Used for extracting the phrase options).
TestFile = 
Please define the style of outputing the sentence phrase options: If 1, store all sentence options first then output them at once, use large memory but faster; if 0, collect phrase options for one sentence and output, use less memory but slower.
batchOutputLabel = 1 
======================================================================================


Please fill in the parameters for seting up the structured learning framework:
======================================================================================
The maximum length of the phrases extracted (e.g. 7 as used in MOSES).
maxPhraseLength = 7
The class setup of the structured learning framework (i.e. 3 or 5).
classSetup = 3
Prune the phrase pairs whose reordering distance (d) is greater than a const distance (distCut).
distCut = 15
The maximum length of the ngram features in the ngram dictionary (e.g. usually choose 3 or 4).
maxNgramSize = 4
The window size around the source phrases (for extracting the ngram features, usually choose 3 or 4).
windowSize = 3
Prune the ngram features that occur less than the minimum number of occurance (minPrune).
minPrune = 1
Prune the cluster that has examples less than the minimum number of training examples (minTrainingExample).
minTrainingExample = 10
The maximum iteration for the perceptron-based structured learning (please refer to Ni et al., 2009) algorithm.
maxRound = 500
The step size for the perceptron-based structured learning (please refer to Ni et al., 2009) algorithm.
step = 0.05
The error tolerance for the perceptron-based structured learning (please refer to Ni et al., 2009) algorithm.
eTol = 0.001
======================================================================================


