Configuration

Command	Explanation
Grammar
ignorePunctuation=false	Includes/Ignores punctuation from data
TAGSET=src/main/resources/english.pos.map	POS tag file location
tagType=Fine	Parse with {Coarse,Fine,Universal,Induced} tags
hasUniversalTags=true	Add column for NAACL Shared Task input/output
NF=Full	Parse with NF: {Full,Full_noPunct,Eisner,Eisner_Orig,None}
trainingRegimen=[readTrainingFiles, HDPArgumentModel, IO, Test]	Operations for experiment
typeRaising=false	Allow TypeRaising
lexTROnly=false	Restrict TypeRaising to lexical items
allowXbXbX=false	Allow for (X/X)\X and (X\X)/X
Grammar Induction
uniformPrior=false	EM init w/ Uniform Trees
maxArity=3	Maximum lexical arity
maxModArity=2	Maximum lexical arity for Modifiers
induceValidOnly=true
complexArgs=false	Allow complex arguments
complexTOP=false	Allow TOP to complex arguments
ALPHA_SCHEME=false	Should hyper-parameters be used as constants or X^a schemes?
alphaPower=[1000.0, 1000.0, 1000.0]	variational hyperparameter
discount=0.0	PY Discount factor 0 <= d < 1
typeChangingRules=null	Special Unary Type-Changing rules
Tagset Induction
BMMMClusters=null	New BMMM Tag mapping
Training
source=supervised	Training setups: induction, supervised
viterbi=false
maxItr=2000	Max # of EM/BW Iterations
threshold=0.01	EM/BW convergence threshold
trainK=1	TopK parses to be computed during training
smallRule=-25.0	Minimum prob/val allowed for a rule
NumClusters=45	Number of clusters to induce with HMM
Training Data
trainFile=[english.AUTO.example]	Training file(s), comma delimited
shortestSentence=1	Shortest sentence to consider
longestSentence=200	Longest sentence to consider
Misc
Folder=ExperimentOutput	Folder for output files
saveModelFile=ExperimentOutput2//Model	file to write the model
loadModelFile=ExperimentOutput2/Model0	file to read the model
savedLexicon=Lexicon.txt.gz	Lexicon to load
CondProb_threshold=0.01	Threshold for discarding categories based on cond prob
threadCount=2	Number of threads to use
api_key=key.txt	API Key for push notification from notifymyandroid.com
Induction's Lexical Learning
lexFreq=5.0	# or percentage of words to learn
nounFreq=0.0	# or percentage of nouns to learn
verbFreq=0.0	# or percentage of verbs to learn
funcFreq=0.0	# or percentage of function words to learn
Testing
testFile=[english.AUTO.example]	List of test files ( comma delimited )
TEX_LANGUAGE=other	Are the sentences in chinese or other?
CONLL_DEPENDENCIES=CC_X1___CC_X2	Whether CoNLL style viterbi parses should be printed and how conjunction should be treated
longestTestSentence=100	Longest allowable test sentence (else: right branch)
testK=1	Number of parses to produces at test time
AUTO_TYPE=CCGBANK	CCGBANK vs CANDC auto files
AUTO Conversion
AUTOFileToConvert=null	Input file [1 auto per line] to convert
ConvertAUTO=TEX	Format to convert AUTOs to
Knowledge Graph
hardBracketConstraints=false	Use Hard Entity constraints when parsing
hardEntityNConstraints=false	Use Hard Entity as N constraints when parsing
softBracketConstraints=true	Use Soft Entity constraints when scoring
softEntityNConstraints=false	Use Soft Entity as N constraints when scoring
softBracketWeighting=0.9	1-Penalty for violating entity constraints when scoring